Wind Shear Alert System

1. Field of the Invention

The present invention relates generally to methods for detecting horizontal wind shear within a geographic location of an airport bounded by a network of sensors located at specific locations. More particularly, the invention relates to wind shear detection systems Which incorporate a network of anemometers to sense wind velocity, speed and direction, near the airport.

2. Description of the Prior Art

The Federal Aviation Administration (FAA) has previously deployed a Low Level Wind Shear Alert System (LLWAS). The LLWAS used sensors or anemometers on towers located at specific locations about an airport. The anemometers measured wind velocity. The original LLWAS system gave information to the air traffic controllers, which information was relayed by radio to the pilot of an arriving or departing aircraft.

The original LLWAS system used a vector difference algorithm to detect horizontal wind shear, i.e., wind shears which affected head wind relative to an aircraft. The LLWAS system issued general area messages about wind shear and was not geographic specific within the wind field about the airport. A wind field is made up of wind velocity vectors for each sensor station over the geographic area in which a network of sensors is established. The LLWAS did not issue runway specific messages and did not estimate the intensity of the wind shear hazard.

A follow-up system to LLWAS, called a Phase II system, has been developed by the FAA to upgrade LLWAS. The Phase II system uses the same anemometers to provide wind velocity and direction data. The Phase II system is partially adapted to handle data noise. The Phase II system has a limited ability to detect wind field divergence, a derived value based on spacial derivatives of the wind field. Specifically, a positive divergence might indicate a microburst wind shear, or a less intense wind shear. Divergence is associated with a loss of head wind, and resulting loss of aircraft performance on takeoff or landing.

Even though the Phase II system attempts to detect wind field divergence, neither LLWAS nor the Phase II system is capable of issuing microburst alarms, as differentiated from any other wind shear. The Phase II system does not give runway specific information about the intensity or location of any wind shear, other than a general regional statement. The Phase II system is the wind shear alert system operating throughout the United States at the present time.

Both the original LLWAS and the Phase II system use the same basic approach. Anemometer data is gathered and processed for the network. A simple wind shear alert, without differentiating between a microburst and other wind shear, is given for a general location. The locations are directional, i.e., north or east. The magnitude of the wind shear hazard, i.e., whether a microburst is present, and its location are not addressed in the prior systems.

Remote sensor stations spaced at specific locations about an airport location have tower mounted anemometers for sensing wind velocity, speed and direction. A wind velocity representation is transmitted to a base station computer which stores U and V values, expressed in meters per second. The U and V values, coordinates of a Cartesian coordinate system, are representative of a wind velocity vector at each sensor station location.

The wind shear alert system processes the wind velocity measurements over the wind field, wind field measurements, which are stored as the wind velocity vectors U and V, to detect wind shear. Wind velocity data for all the sensor stations, and preselected parameters from an airport configuration file, are combined and presented to the air traffic controller as runway specific alphanumeric messages concerning microbursts and wind shear alerts.

The first step in the wind shear alert system process is to enhance data quality. Missing data are adjusted for both long term and intermittent gaps. Various filtering techniques are applied to the wind field measurement data to suppress short term variations and localized meteorological variation. Data availability flags are set and used to guide the analysis of the remainder of the steps of the system process.

A robust network mean estimate is computed using data trimming and temporal averaging. The network mean computation yields an estimate of the mean wind field vector, or ambient wind, and the noise level in the data. The mean wind field vector is used to detect evidence of wind shear beyond the network. The noise level in the data is used to adjust thresholds to adapt to volatile wind conditions.

A divergence analysis is used to detect primary wind shear hazards, microbursts and gust fronts. Microbursts are characterized by positive wind field divergence. Gust fronts show strong negative divergence or convergence.

The divergence analysis is based on a family of edges and triangles whose vertices are at selected sensor stations. An edge is a line segment that joins a pair of closely positioned sensor stations, typically 1.5 to 5.5 Km. A triangle is formed by a triple of closely positions sensor stations. A numerical derivative technique is used to measure the divergence on each edge and triangle. If a divergence is detected, then a microburst head wind loss estimate is generated, based on a simplified symmetric microburst model. This information is the basis for a microburst alert message.

If a weaker divergence or a convergence is detected, then the runway oriented head wind loss and gain are computed for each runway. Wind vector components parallel to the runway are compared to establish the loss or gain for each runway. A wind shear message is issued for each runway that has a persistent microburst level divergence or a persistent loss or gain in excess of 15 knots. The location of the wind shear along the runway is identified.

The wind shear alert system of the present invention supplies alarms which are transmitted to the air traffic controller at the particular airport. The alarms indicate what type of wind shear is present, Which runways are affected, and which wind shear is most hazardous when multiple wind shears are present. Alert types include a microburst alert (MBA), wind shear alert with head wind loss (WSA/loss), and wind shear alert with head wind gain (WSA/gain). The magnitude of the loss or gain and the location of the wind shear along the runway is computed. An additional test is executed to determine if there is evidence of wind shear beyond the network. This information is then radioed to the pilot of an aircraft using the specific runway.

The message generated as a result of the alarm for each runway therefore includes the information of alert type, gain or loss value, loss being a negative, and shear location, given a value of 0, 1, 2 or 3. The shear location values indicate nautical miles from the threshold end of the runway, 0 being on the runway, 1 being within one mile of the runway, 2 being between one and two miles of the runway and 3 indicating wind shear between two and three miles off the end of the runway. Messages are issued for each arrival runway and for each departure runway.

Other aspects, features and details of the present invention can be more completely understood by reference of the following detailed description of a preferred embodiment, taken in conjunction with the drawings, and from the appended claims.

FIG. 1 is a schematic illustration of components of a wind shear alert system in accordance with the present invention.

FIG. 2 is a schematic plan view of runways of an airport employing the wind shear alert system illustrating an exemplary location of some of the components shown in FIG. 1.

FIG. 3 is a flow chart showing the steps of the process of the wind shear alert system in accordance with the invention.

FIG. 4 is an illustration of a Cartesian coordinate system showing a wind vector at a sensor station employed in the system shown in FIG. 3.

FIG. 5 is a schematic elevation illustration of a microburst and a pair of sensor stations, defining an edge, along a runway, an aircraft shown on final approach to a runway of an airport such as shown in FIG. 2.

FIG. 6 is a perspective illustration of a microburst and three sensor stations, defining a triangle, and an aircraft shown on final approach to a runway of an airport such as shown in FIG. 2.

FIG. 7 is a plan illustration of a runway and triangle showing the physical relationship in detecting a microburst head wind loss and a runway oriented head wind loss using the system shown in FIG. 1.

The present invention detects horizontal wind shears and particularly detects and differentiates between hazardous wind shears, or microbursts, and other wind shears. Microbursts are downwardly directed air flows, which impact the ground perpendicularly, in which case they are symmetrical as shown in FIG. 5, or obliquely, in which case they are asymmetrical, resulting in a noncongruous wind field near the ground. A microburst typically is one and a half to five kilometers in radius at the ground. Horizontal wind shears 18 (FIGS. 5 and 6) are generated from the ground to an altitude of 1-2 kilometers. These wind shears 18 appear to an aircraft 20 arriving or departing from a runway 24 as a head wind gain or a head wind loss.

As seen in FIGS. 5 and 6, a head wind gain is followed by a head wind loss, after the aircraft 20 passes through or beneath the microburst 16. The change in head wind from a gain to a loss results in a hazardous situation as the pilot of the aircraft 20 may initially bleed off airspeed as a head wind gain is encountered and then must increase airspeed, as he experiences a head wind loss, in order to avoid stalling the aircraft 20. The situation is aggravated by the fact that the aircraft 20 is either on final approach or just departing the runway 24 at an airport 26, and is at very low altitude. The adjustments in airspeed to compensate for the head wind losses and gains are difficult for the pilot to execute when there is no warning of the possible hazardous condition.

A wind shear alert system 10 of the present invention, shown in FIG. 1, uses a network 12 of remote sensor stations 14 to detect a microburst 16 (FIGS. 5 and 6). The remote stations 14 of the network 12, shown in FIG. 2, supply wind field measurements or data 36 (FIG. 3) to a base station computer 34 which stores the data as wind vectors 28 (FIG. 4) representative of the wind velocity, speed and direction. The wind field is represented by the wind vectors 28 of each of the stations 14 over the network 12. The wind shear alert system 10 then processes the wind field data 36. The base station computer 34 generates alphanumeric alarms 48 in the form of messages differentiating between microbursts and other horizontal wind shear. Besides differentiating the wind shear type, i.e., microburst or other wind shear, the alarm 48 also issues messages capable of distinguishing between a microburst or wind shear alert with head wind gain or head wind loss. The gain or loss value and the microburst or shear location in nautical miles from the threshold of the specific runway 24 of the airport 26 is indicated in the alarm message.

The alarm messages are of value to pilots of the aircraft 20 during arrival or departure from the runway 24 of the airport 26. The pilots have advance warning of the type of wind shear, microburst or other wind shear, the magnitude of the wind shear in terms of relative gain or loss of aircraft performance based on head wind, and the location of the wind shear in nautical miles from the runway. Based on this information, the pilot is forewarned and is prepared to make adjustments in his airspeed to compensate for the wind shear losses or gains in the aircraft 20 performance. In executing an arrival, in extreme situations the pilot may elect to go around the airport 26 and try again after the microburst passes.

The outline for a computer program implementing the process of the wind shear alert system 10 is shown in FIG. 3 and Appendix I; a specification for a computer program implementing the wind shear alert system 10. With reference to Appendix I, page ii, the process implementing the wind shear alert system 10 and shown in the flow chart of FIG. 3 consists of the steps of filtering wind data 38, network mean computation 40, divergence 42, loss or gain 44, location 46, and alarm messages 48. Still with reference to Appendix I, pages iii through xxxiv show additional detail of each step of a computer program algorithm that would implement the method of the wind shear alert system 10.

The wind shear alert system 10 of the present invention is able to generate the alarm messages based upon the wind field measurements or data 36 and parameters which are preselected and stored in the base station computer 34 in an airport configuration file 30, attached hereto as Appendix II. This input is then processed as shown in FIG. 3 and Appendix I.

Representative constants and the computations necessary to generate the parameters used in the airport configuration file 30 are shown in Appendix III. Appendices I, II and III are attached hereto and incorporated herein by reference.

The remote sensor stations 14, which define the network 12 shown in relation to the runways 24 in FIG. 2, are positioned strategically relative to the runways 24. The exact locations of the sensor stations 14 will vary with each different airport 26, dependent upon the runway location and other geometric and geographic considerations. Generally speaking, the sensor stations 14 are located between 1.5 and 2.5 kilometers apart and generally parallel to and on either side of a single or parallel extending runways 24.

Each of the remote sensor stations 14 includes an anemometer 32 mounted on a tower (not shown). The anemometer 32 measures wind velocity, speed and direction. Wind speed and direction, represented as the wind vector 28, having U and V components in a Cartesian coordinate system shown in FIG. 4, are generated. A signal representative of the wind vector 28 is transmitted by a telecommunications package 33 associated with the remote sensor station 14 to the base station computer 34. Additional details of the wind shear alert system 10 can be obtained by reference the United States Department of Transportation's FAA Specification on Low-Level Wind Shear Alert System (LLWAS), Technical Specification Number FAA-E-2697A, dated Mar. 6, 1989, which specification is incorporated by reference herein as though fully set forth.

The wind field measurements 36 are periodically polled by the base station computer 34 (FIG. 3). The results of a predetermined number of polling cycles are maintained in the base station computer 34 and are utilized in performing the wind shear alert system computations as will be discussed hereinafter. The base station computer 34 stores the airport configuration file 30 and the wind field measurements 36. In addition, the filtering of wind data 38, the network mean computation 40, the divergence computation 42, the loss gain computation 44 and location of any wind shear 46 are all computed by the base station computer 34 using a computer program based on the algorithm of Appendix I. Based on the results of those computations, the alarm messages are issued to an air traffic control tower 27 at the airport 26. At the tower 27 the message is displayed on an air traffic controller display 49. The air traffic controller radios the information to the pilot of the aircraft 20.(FIG. 1.)

Referring to FIG. 3 and to Appendix I, the first step of the process of the wind shear alert system 10 is to filter the wind data 38. Pre-checks are made to insure that good data on the wind vectors 28 is available in the sense that the sensor stations 14 are reporting current winds. The data are filtered through a weighted mean running average process, which smooths the data. The parameters for controlling the weighted mean running average process are the length of the filter and the weightings that are used in the computation of the weighted means. The weighted running mean is based on holding the last several data values for each station in a circular queue and computing the filtered value (u(k), v(k)) of the station wind velocity vector at poll number k; ##EQU1## where w(0), . . . , w(n-1) is the list of filter weights. Further details of a computer based calculation are given in the process FILTER.sub.-- WIND.sub.-- DATA of Appendix I, p. vi and vii.

During the computation of the network mean computation 40, a sample network mean, a wind vector representative of the ambient wind 22 in the network 12, is estimated for each poll taken of the wind field measurements 36. This estimate is computed based upon the filtered wind data 38. During this step of the method of the wind shear alert system 10, it is determined whether or not some of the remote sensor station 14 data are extreme outliers. If they are outliers, they are discarded and the current network mean estimate is computed from the remaining data. The standard deviation is also computed. These computations are compared against the network mean and standard deviation from the previous polls by weighted averaging. This combination of averaging across the network 12 and averaging over time creates very stable estimates.

The network mean estimate is computed by

U.sub.-- temp=1/m .SIGMA. U(j)

V.sub.-- temp=1/m .SIGMA. V(j)

U.sub.-- bar=A * U.sub.-- bar(previous)+(1-A) * U.sub.-- temp

V.sub.-- bar=A * V.sub.-- bar(previous)+(1-A) * V.sub.-- temp

where m is the number of untrimmed stations and the index j runs over all untrimmed stations. Details are provided for a computer program implementing the process in NETWORK.sub.-- MEAN of Appendix I, p.p. ix through xi.

The next step of the method of the wind shear alert system is the computation of the wind field divergence 42 and the determination of whether or not there is a persistent wind field divergence or wind sheer. This determination is based on the filtered wind data 38, as well as parameters from the airport configuration file 30 relating to the actual geometry of the network 12. Divergence is computed along edges 50, and over triangles 51 (FIGS. 2, 5, 6 and 7). Edges 50 are visualized as the line segment between two relatively close pairs of remote sensor stations 14. These pairs of stations are listed in the airport configuration file 30. The triangles 51, consisting of a triple of the remote sensor stations 14, allow computation of a two dimensional divergence. Again the remote sensor stations 14 of the triangles 51 are relatively close and are listed in the airport configuration file 30. The airport configuration file 30 also contains the direction vectors of the edges 50, direction vectors of the sides of the triangles 52, the lengths of the edges 50 and the areas of the triangles 52. See Appendix II.

The wind field divergence 42 may be visualized physically by reference to FIGS. 5 and 6. As seen in FIG. 5, a ambient wind 22 is perceived by the arriving aircraft 20. The microburst 16 strikes the ground intermediate two sensor stations 14, creating wind shears 18. The aircraft 20 is on approach to the airport 26 to land on the runway 24. The remote stations 14 will indicate the wind velocity vectors 28 in essentially opposite directions for the shown stations 10, indicating the presence of the wind shears 18. As previously stated, the problem for the pilot of the aircraft 20 is an initial increase in head wind at one end of the edge 50 followed by a decrease in head wind as the pilot passes through the microburst 16.

With reference to FIG. 6, three remote sensor stations 14 define the triangle 52. Rather than flying along the edge 50 parallel to the runway 24, as was the case in the example of FIG. 5, in FIG. 6 the aircraft 20 passes through the ambient wind 22 and travels over the triangle 52 defined strategically relative to the runway 24. The head wind loss or gain is analyzed in two dimensions rather than one, as was the case in the divergence analysis for the edge 50.

Turning to FIG. 2, the network 12 of the remote sensor stations 14 can include numerous edges 50 and triangles 52 as are indicated in dotted line. Depending upon which of the runways 24 an arriving or departing aircraft 20 uses, different ones of the edges 50 or the triangles 52 may be more important to the wind shear alert system 10. One of the edges 50 can also be part of one of the triangles 52.

Divergence is compared to divergence thresholds. The ratio of the measured divergence value is made to a divergence threshold value. A ratio greater than 1 indicates that a hazardous divergence wind shear is occurring at the indicated edge 50 or triangle 52. The airport configuration file 30 contains triangle divergence thresholds and edge divergence thresholds, as well as triangle and edge convergence thresholds used to measure whether or not the severe negative divergence is occurring, resulting in a head wind gain.

A noise adaptive threshold, is computed as part of computing the divergence 42. The adaptive threshold is a factor multiplied by the standard deviation computed from the network mean computation 40. In the case of extremely gusty winds, the standard deviation will be a large number and the noise adaptive threshold will be used to desensitize the wind shear alert system 10 from issuing alarm messages that are not the consequence of hazardous wind shear but rather just expected variations in a very noisy wind field.

If, during the computations of the divergence 42, an edge 50 or triangle 52 is determined to have a large divergence or convergence, then it is necessary to determine whether or not that situation has persisted. Noisy wind fields can have fluctuations over time which appear and disappear and do not persist. Only persistent fluctuations indicate a hazardous condition to the arriving or departing aircraft 20. Again, parameters from the airport configuration file 30 control the determination of whether or not a divergence or convergence situation has persisted.

The edge divergence is the vector dot product

where (delta.sub.-- U,delta.sub.-- V) is the wind differential along the edge, (dir.sub.-- x,dir.sub.-- y) is the edge direction vector and L is the edge length, both from the airport configuration file 30, Appendix II.

The triangle divergence is given by

where the numerical derivative estimates (Ux, Uy) and (Vx, Vy) are computed by solving the matrix systems ##EQU2## where the values U1, U2, U3, V1, V2, V3 are the wind field values at the triangle vertices and the values X1, Y1, X2, Y2 are the triangle side vectors, from the airport configuration file 30, Appendix II. Details for a computer program are provided in the process DIVERGENCE of Appendix I, p.p. xii and xiii.

Once a divergence is detected at the airport 26 and within the network 12, the method of the invention analyzes the loss 44 (FIG. 3) associated with the microburst or wind shear. A scaling factor, the effective length, is multiplied by the value of the divergence to give an estimate of the loss that an aircraft 20 could experience if it flew through the center of the microburst. The derivation of the effective length is based on a statistical analysis involving a theoretical microburst model, the lengths of the edges 50, and the sizes and shapes of the triangles 51. The loss estimate is proportional to the divergence measured in the network 12. If the divergence value is doubled, then the loss of head wind estimate is also doubled.

If a microburst loss of head wind is not detected, then a wind shear condition of less intensity might be present. Weaker divergence or convergence coupled with runway oriented loss or gain is indicative of wind shear 18. The runway oriented loss and gain is determined by station-to-station comparison. Remote sensor stations 14 that are relatively near each other along the runway 24 (FIG. 7) are used for the loss or gain computation. The list of those pairs are stored in the airport configuration file 30 for each runway 24, which also includes the along runway distance between the pairs of the stations 14.

After the divergence, convergence and loss or gain computations have been made, a check is made to see if a persistent loss or gain situation exists along the runway 24. The basis for an alarm message is a persistent divergence or persistent convergence on some edge or triangle or a persistent loss or gain on some runway. Persistence is measured in the sense of at least n satisfactory occurrences out of the last m polls. Circular persistence queues maintain a record of whether the divergence or convergence was in excess of the threshold value for each of the previous m polls, for each triangle and edge. Circular persistence queues also record if there was a loss or gain situation on each runway in excess of the wind shear threshold for each of the m previous polls. A count of the successes determines whether each persistence criteria is met. The lengths of the persistence queues and the persistence thresholds are provided in the airport configuration file 30, Appendix II. This approach to the handling of persistence allows a partial separation of the timeliness considerations inherent in the running average filters and additional delays that might be desired for alarm stability.

The final step is a decision as to whether or not to issue the alarm 48 and what type of alarm message to give must be generated. Several tests are made. Initially a divergence alarm indicates whether or not there is a microburst. Depending upon which edge 50 or triangle 52 is involved in the divergence alarm, locations for an arrival and departure aircraft 20 are determined. The location comes from a list of pointers available from the airport configuration file 30. A convergence alarm test is also done, but instead of looking for a positive divergence value, a negative divergence value is involved.

Wind shear with loss and wind shear with gain alarm tests compare the loss and gain values with the wind shear threshold set in the airport configuration file 30. There are also microburst tests. A divergence test leads to a microburst alarm situation if the loss value is large compared with the microburst threshold. The microburst threshold also comes from the airport configuration file 30. The alarm 48 step also arbitrates which alarm is most hazardous when several alarms are simultaneously available on the runway.

The basis for alarm arbitration is that a microburst is more hazardous than any other wind shear and that if the magnitudes are nearly equal, then a loss is more hazardous than a gain. Therefore, in the case of multiple alerts, if one of them is a microburst alert, then that alert, loss, and location is sent forward to the alphanumeric displays. In the absence of a microburst alert, if there is both a loss and gain, then the loss is given, unless the gain is substantially larger. Parameters from the airport configuration file 30 (LOSS.sub.-- INCREMENT and LOSS.sub.-- BUFFER) control the latter decision.

The alarm step of the method also predicts whether or not there is a possibility of a wind shear outside the network 12. A computation is made using the runway remote sensor station 14 locations and the runway direction vector. These values are taken from the airport configuration file 30, Appendix II.

As shown in FIG. 3, the airport configuration file 30 supplies a variety of constants to the computer program generated pursuant to the algorithm disclosed in Appendix I. The parameters for the airport configuration file, shown in a source code notation in one column and a short technical description in a second column are listed in Appendix II attached hereto and incorporated herein. The airport configuration file contains numbers which are specific to the airport 26 at which the wind shear alert system 10 is installed. For example, specific numbers include the latitude and longitude of specific runways and the location of the sensor stations 14.

Appendix III, which is attached hereto and incorporated herein, shows typical constants for some of the parameters of the airport configuration file 30 and the methodology in computing other parameters. These constants are used to control the operation of the computer program and include flag values, tolerances and count limits. In addition, there are filter constants which are applicable to several levels of filtering in the computer program for generating the airport configuration file 30. The wind field measurements 36 are filtered by weighted running means and the network mean and standard deviation are filtered by single poll recursive filters. The lengths of these filters and the filter constants are tailored to the length of the polling cycle and desired algorithm performance. Also included in the constants and computations of the airport configuration file is the network geometry, the locations of sensor stations 14, edges 50 and triangles 52, shown in FIGS. 5, 6 and 7.

The airport configuration file 30 will now be discussed in detail. The airport configuration file 30 contains lists of edges 50 and triangles 52 that are used in the wind field divergence computations. The sensor stations 14 are listed in some order and numbered. Although not required for algorithm execution, the latitudes and longitudes of the station 14 locations are included. Each edge 50 is described by the pair of station numbers that describes its endpoints, by its length, and by its unit direction vector. Typically, a pair of sensors 14 defines an edge 50 if its length is between one and a half and five and a half Km. Each triangle is described by the triple of station numbers that describes its vertices, by its area, and by a pair of unit direction vectors for two of its sides. The triple of station numbers is listed to correspond and the pair of unit direction vectors share the first vertex as their common base point. Typically, a triple of sensors defines a triangle if each of the sides has its length between one and a half and five and a half Km and if the smallest angle of the triangle is larger than twenty-five degrees.

Divergence and convergence thresholds are partially based on statistical analysis and are designed to avoid issuing alerts on statistically expected divergence values. In addition, these thresholds and the persistence thresholds incorporate management requirements for system performance such as minimum hazard level and timeliness of the alert.

Effective lengths are factors used to convert divergence estimates from triangles 52 and edges 50 into loss estimates when there is evidence that a microburst is present. They have been chosen statistically to give an unbiased estimate of the head wind loss that the aircraft 20 might encounter if it were to enter the microburst.

The loss or gain of head wind for each of the runways 24 needs lists of pairs of sensor stations 14 in the airport configuration file 30. For each pair of remote sensor stations 14, the station vertices and the along runway distance between the sensor stations 14 is listed in kilometers.

Each station sheltering sector is a wedge or arc of directions from which sheltering by some physical feature of the airport 26, such as a hill or building, renders the sensor station 14 to have unreliable wind vector 28 measurements. A wedge is provided for each sensor station 14. The wedge consists of the initial angle and the final angle, in degrees and measured counterclockwise from magnetic north. If there is no sheltering, then the wedge is described as 0, 0.

Runway geometry of each of the arrival runways 24 is represented by a unit direction vector. In addition, the runways 24 are described by their latitudes and longitudes and by their Cartesian coordinates in the airport coordinate system. This information is not necessary for the algorithm but is included for completeness.

Finally, runway alarm location pointers are generated by the airport configuration file 30. Each sensor station 14, edge 50 and triangle 52 is associated with two positions along each of the arrival runways 24, the arrival location and a departure location, the latter for the oppositely directed runway 24. The location pointers have values of 0, 1, 2, 3 or RWY-LOC FLAG. A zero (0) indicates "on the runway," 1, 2, or 3 indicates the distance from the runway 24 threshold in nautical miles and RWY-LOC FLAG indicates that the element is not associated with this runway.

In operation, the process includes wind data filtering 38 by a running weighted mean. The wind field measurement 36 data from the current poll of the remote sensor stations 14 and several previous polls, approximately 2 minutes worth of data kept in circular data queues, is combined with filter weights to compute the weighted mean of the wind field measurements 36 from the current poll of the remote stations 14 and whatever number of previous polls is appropriate. The advantage of using this filter mode is that it is easy to determine what happens to the data, both from the viewpoint of suppressing the variance of the data and from the viewpoint of the amount of time lag if an impulse or step function is introduced into the data stream. Separate filters are used for the four basic computations, i.e., Mode 1 is the network mean, Mode 2 is divergence, Mode 3 is microburst loss and Mode 4 is runway oriented loss and gain.

The network mean computation is used to get a robust estimate of the mean. Data trimming is based on chi-squared tests. After the removal of outliers, estimates of the network mean and standard deviation are computed from the current poll. These estimates are combined with the estimates from prior polls by weighted averaging. The divergence estimation involves numerical differentiation along the edges 50 and the triangles 52. The triangle derivations are combined to compute the two-dimensional divergence. The numerical differentiation process is a simple first order derivative procedure that is standard in numerical mathematics.

When there is evidence of persistent divergence on a edge 50 or triangle 52, then the value of that divergence is re-estimated using more heavily filtered wind field measurements 36, Mode 3. This stabilized divergence is multiplied by the effective length, a scaling factor obtained from the airport configuration file 30 to provide the microburst lost estimate. If the microburst estimate is less than the microburst threshold, only wind shear is indicated and the runway oriented loss or gain is computed.

The loss or gain estimates are computed by simply looking at the component of the wind field and the direction parallel to the runway 24 and going a station-to-station comparison to determine whether or not a loss situation or a gain situation is available. These stations are listed in the airport configuration file 30.

In both circumstances, loss or gain, the location of the most severe loss or gain is estimated and that is used to estimate the location at which the wind shear 18 will be encountered by an arriving or departing aircraft 20. In the hierarchy of the alarm tests 48 evidence of a microburst alert, is always given to the pilot. A weak divergence alarm, possibly not rising to the level of a microburst, is compared to the runway oriented loss or gain and a decision is made weather or not the loss or gain situation is predominant. The dominant hazard situation, loss or gain, is radioed to the pilot.

The alarm 48 concerning wind shear 18 outside the network 12 is issued only in a case where there is a gain situation on the runway 24, which is at the far end of the portion of the runway that is protected by the network. If there is a gain situation, it is possible there is more wind shear 18 beyond the network 12. The method evaluates the situation and issues an alarm if it is appropriate.

Although the present invention has been described with a certain degree of particularity, it is understood that the present disclosure has been made by way of example, and changes in detail or structure may be made without departing from the spirit of the invention, as defined in the appended claims. ##SPC1##