theoretical cross section (m 2 ) X S L

Transcription

1 EFFECTS OF BIRDS ON RADAR TRACKING SYSTEMS J R Moon J R & J Moon, UK INTRODUCTION Tracks from birds, or flocks of birds, present a problem to designers of land and sea-based radar tracking systems, since the users of these systems are generally uninterested in such tracks and would prefer them to be removed from the synthetic track picture. The difficulties are compounded by the fact that users, while wishing to remain ignorant of bird tracks, would still like to be informed of tracks of other fairly slow flying objects such as light aircraft and helicopters. In naval or coastal radar systems it is usually also a requirement to be able to form tracks on moving ships and boats of all descriptions, some of which can move at similar speeds to birds. For operational reasons there is a trend towards radars being made more sensitive, and a side effect of this is to increase the chances of birds forming a significant proportion of all tracked objects. This paper examines some of the characteristics of bird detection and flight and analyses examples of surveillance radar recordings affected by bird tracks. RADAR CROSS SECTIONS OF BIRDS The radar reflectivity of a bird is predominantly due to the 65% of its mass that is water, and an approximate model of the radar cross section can be obtained by regarding the bird as a water sphere of equivalent mass to the water content. The solution to Maxwell s equations for a conducting sphere [Barton (1)] shows that the ratio radar cross section circumference varies with physical cross section wavelength as in figure 1. Figure 1 Normalised radar cross section for a conducting sphere and equivalent birds It can be seen that for many birds and wavelengths the scattering will come within the resonance region and the cross section can be expected to vary by up to a factor of 3.5 (5.4dB) from the optical cross section. However, a general correction factor of approximately 0.56 is also needed for the reduced reflectivity of water compared with a perfect conducter [Eastwood (2)]. Table 1 below shows the computed cross sections (in both m 2 and dbsm) for various generic bird types and for three different radar wavelengths (3cm, 10cm and 23cm) corresponding to X, S and L band frequencies. generic type bird mass (g) theoretical cross section (m 2 ) theoretical cross section (dbsm) experimental cross section (dbsm) warbler sparrow starling plover pigeon duck swan Table 1 Theoretical and experimental cross sections for various bird types

2 The experimental data in the table are taken from Nathanson (3) and perhaps indicate a slight tendency for this model to overestimate cross sections of these bird types although individual birds can clearly be expected to vary in weight from the indicative values given. Nathanson also gives data that indicate that the cross section can vary by up to 9dB depending on bird aspect, so the above sphere model is clearly approximate. Another limitation is in the Rayleigh scattering region where scattering depends more on volume than on mass (1), so for very small birds at long wavelengths the model may give a compensating underestimate of the cross section. Except for very large birds the cross sections are small compared with those of aircraft, ships or missiles. However, when birds are in flocks there may be multiple birds within a resolution cell of the radar and birds at altitude may be viewed against a clutter-free background which will also enhance the chances of detection. It is commonly supposed that the stochastic variation in the reflected signal from a single bird is given by a log normal distribution [e.g. (1), (3)]. BIRD FLIGHT CHARACTERISTICS The available data on bird flight characteristics affecting radar are covered in detail in Moon (4) but some of the main features are summarised below. Bird flight speeds Reliably measured bird ground speeds vary from a few ms -1 up to a maximum of 43ms -1 and would seem to depend on a number of factors, notably: Size The broad trend is for speed to increase with size, but there are also many exceptions. Shape Birds with large wings and relatively low weight, such as gulls, are very manoeuvrable but generally slow; on the other hand birds with a high wing loading, such as ducks and geese, are less manoeuvrable but can maintain good straight line speed. Activity Birds on migration, or on transit from roosting to feeding areas, tend to fly faster and with more constant course than the same species on local foraging flights. Effects of wind These effects seem complicated and cannot be predicted simply by vectorally adding the wind speed to the bird speed measured with zero wind. Bird density Close to coastal roost sites there could be 10 5 or 10 6 birds within 50km of a ship-borne surveillance radar, and similar densities could occur within well-vegetated areas on land although, in general, only a small fraction of these are likely to be airborne at a given time. An exception is at times of peak migration when there could be more than 10 6 birds in the air within a 50km radius. However many of these birds will be in flocks, and so the notional number of potential radar tracks could be two orders of magnitude lower than this figure and could be further limited by the action of the CFAR system. EXAMPLES OF BIRD TRACKS AFFECTING TRACK PICTURES This section shows data from several recordings made at sea with a ship-borne surveillance radar in different conditions between 1993 and None of the recordings are typical, in the sense of being everyday examples of how birds affect radar pictures, but on the other hand there is no reason to believe that the recordings represent highly improbable extremes. Example 1 Figure 2 shows tracks attributable to birds recorded during a 40 minute period off the Welsh coast from a ship-board S-band radar operating in very calm conditions. The associated processing system automatically formed tracks from coherent sequences of plots and retrospective analysis programs were used to categorise tracks by origin and to estimate their speed. From the point of view of the operators, the bird tracks were an unwelcome source of clutter in the picture. Bird altitude during flight Around Europe the vast majority of migrating birds fly below 2000m [Alerstam (5)} but birds on transoceanic migrations often fly up to 6000m [Berthold (6)], and there are also many examples of flocks of swans and geese being observed at altitudes up to 9000m. The highest recorded bird is a vulture at 11,300m. Figure 2 Tracks from birds in very calm conditions inside 75km and (inset) 10km

3 Altogether 400 tracks attributable to birds were recorded during a 40 minute period and it can be seen that there are two distinct sets. The fairly dense set of 172 tracks in an annular cluster close to the ship (see inset picture) is characterised by low speeds (5ms -1 to 10ms -1 ) and rather erratic flight patterns and is attributable to birds on local foraging flights. The tracks at longer ranges have speeds in the range 15ms -1 to 25ms -1 and are due to birds making more direct flights across the surveillance area. The fragmentation of the tracks is due to the mirror-like sea surface giving a very pronounced lobing structure to the radar propagation pattern. The spacing of the interference lobes varies with altitude and range but, by running the radar propagation model CARPET (7), it was deduced that these tracks had elevations variously between 500m and 2000m. At the ranges involved it was likely that the tracks originated from flocks rather than individual birds. Example 2 The bird tracks shown in figure 3 were recorded in January at the western end of the English Channel. The data again originated from an S-band surveillance radar but this time the radar was employing an MTI filter and the sea conditions were fairly rough (sea state 4-5) with a brisk south-westerly wind (mean surface speed 12ms -1 ). The 97 tracks shown had durations between 20 seconds and 6 minutes and were all travelling roughly in the direction of the prevailing wind, although this is not totally apparent from the figure, which is essentially a plot of the motion of the tracks relative to the ship which made a manoeuvre part way through the 28 minute recording. When the effects of ship motion are removed from the data, the distribution of the ground speeds of the 91 tracks is as shown in the upper part of figure 4, with a mean of 29ms -1 and a maximum of 37ms -1. The standard error of the speed estimates varied with the range and duration of tracks, but averaged 1.5ms -1. The maximum range of detection of any of these tracks was 33km. bird track speed (ms -1 ) bird track speed (ms -1 ) Figure 4 Speed distribution of bird tracks observed in English Channel with S-band (above) and L-band (below) search radars Figure 3 Tracks from birds recorded in English Channel with brisk south-westerly wind Several hours later, when the wind was a little stronger, (mean surface speed 14ms -1 ) a further track recording was made, but on this occasion with an L-band radar on board the same ship. The speed distribution from this recording is shown in the lower part of figure 4. These 57 tracks had an average speed of 27ms -1 and a maximum of 42ms -1. The numbers detected at lower speeds, and hence the average speed, will be a function of the receiver processing, rather than the birds, but the main interest is in the higher end of the distribution. Even taking the prevailing wind into account, these ground speeds are fairly high and indicate that these birds are probably geese or large ducks. Although the fastest speeds during these recordings were close to the maximum of any of the bird speed measurements reviewed in (4), their occurrence indicates that bird speeds of 40ms -1 are more than a remote possibility. Considerations of the radar horizon indicate that some birds at least were well above sea level and the points of entry and exit to and from the conical dead zone centred around the receiver indicate that some tracks had heights of a few thousand metres.

4 Example 3 This third example is a recording made in the North Sea, about 100km east of the Forth and Tay estuaries, in mid-september. The picture of the unprocessed plots received by an L-band radar within 30km of the ship in a 50 minute period is shown in figure 5. It can be seen that the plots are predominantly in coherent streams, attributable to birds, rather than randomly distributed as one would expect from sea clutter. Post trial analysis indicates that approximately 300 plots per scan are attributable to birds, however the probability of detection of individual birds is often well below 1 and the number of birds present and giving rise to plots is probably of the order 500 to 600. track course (deg.) track course (deg.) Figure 6 Distribution of course and speed of bird tracks recorded from L-band (above), and S-band (below) search radars. Figure 5 Plot picture recorded in North Sea off South East Scotland In this scenario, the number of plots per scan that were due to birds considerably exceeded the number of plots due to objects of interest. The minority of the plot sequences that were formed into tracks by the automatic track formation system, and which had more than 10 plots, were subsequently analysed. A plot of the distribution of the course and speed of these birds is shown in the upper half of figure 6, and a similar analysis of tracks recorded from an S- band radar over an overlapping but slightly different time span is shown immediately below. These plots show that there were two distinct populations of birds flying in different directions, with those going north east flying faster than those flying north west. The different proportions of each type in the two scatter plots reflect the different sensitivities of these two radars to birds of different sizes and speeds. The Tay estuary is a well-know roost site for Eider ducks which typically fly around 19ms -1 [Alerstam (8)] and this could be one of the species present. DISCRIMINATION POSSIBILITIES The data given in this paper illustrate that there is considerable overlap in the characteristics of bird tracks and the tracks of interest in many tracking systems. Discrimination against bird tracks might be attempted at various points in the radar processing chain, for example: at the radar signal processing stage, via the use of pulse doppler, MTI or MTD techniques, or via the design of the sensitivity time control (STC) at the plot formation stage, via the observed signal strength and/or the peak frequency of the doppler signal at the track formation stage, via the track-averaged plot strength, the apparent Pd and the apparent track speed and height.

5 The presence of wing beat modulation has also been suggested as a possible discriminant, but this is not a ready proposition for many forms of radar system, particularly as the range of wingbeat frequencies that are likely to be encountered is wide (typically 0-20hz). Depending on the operational aims of the radar system, no single technique or discriminant is likely to be entirely successful at eliminating birds from the track picture. Noyes (9) describes a fuzzy logic method that achieves reasonable discrimination based on a combination of track characteristics, and it would seem that a multi-faceted approach is a necessity if there is also a requirement to be able to form tracks on fairly slowly-moving man-made vehicles. 7. Huizing A G and Theil A, 1993, "CARPET Radar Performance Analysis Software and User's Manual", Version 1.0, Artech House. 8. Alerstam T, Gudmundsson G A, and Larsson B, 1993, "Flight tracks and speeds of antarctic and atlantic seabirds: radar and optical measurements", Philosophical Transactions of the Royal Society of London, series B, 340, Noyes S P, 1998, "Track classification in a naval defence radar using fuzzy logic", IEE Control Division Colloqium on Target Tracking and Data Fusion, Birmingham, IEE Digest No. 98/282. CONCLUSIONS The data and examples given in this paper illustrate that: (a) the kinematic characteristics of birds are often similar to those of some man-made vehicles (b) the observed speeds of birds, and their response to prevailing winds, are such that even if MTI is in use there may be sufficient birds with ground speeds above the MTI threshold to give rise to a significant number of tracks (c) bird tracks, especially in a coastal environment, can form a substantial proportion of the track population of a surveillance radar picture (d) if automatic track initiation and processing is employed, then there is a need to ensure that this processing can cope with the potentially high number of plots due to birds and the high number of tracks, both hypothetical and confirmed, that are consequently formed. Acknowledgement The analysis of the bird tracks in the radar recordings was funded by QinetiQ, Portsdown. REFERENCES 1. Barton D K, 1988, "Modern radar systems analysis", Artech House. 2. Eastwood E, 1967, "Radar ornithology", Metheun. 3. Nathanson F E, 1990, "Radar design principles", 2 nd edition, McGraw-Hill. 4. Moon J R, 2002, "A survey of bird data as it affects radar", Radar 2002, Edinburgh, October. 5. Alerstam T, 1990, "Bird migration", Cambridge University Press. 6. Berthold P, 1993, "Bird Migration: A General Survey", Oxford University Press.