Analysis of the Bentwaters Slow Cluster - Part 3
An Opinion by Martin Shough
e.) CAT, Incoherent Forward Scatter . . . and beyond
Despite the result in Section d. it cannot be denied that certain features of the Track D report remain suggestive of some type of atmospheric propagation anomaly. Perhaps extremely efficient direct power-reflection could occur at incidences near the normal? This would enable returns by first trip from a near-tropopausal structure to display at the real slant range to the layer, or maybe as little as about 9 or 10 miles for the CPN-4's 45-degree cosecanting. This theory would also account for evidence of signal strength diminishing with displayed range since direct backscatter to the radar from the layer would be expected to be more efficient at angles nearer the normal, turning to forward scatter as the elevation dropped significantly below 45°.
But this is still in conflict with the report that the echoes were first picked up at 8 miles and overflew the radar, and in fact time remains a problem as well as distance. Evidence of a possible scattering layer at about 31,000' exists in the Hemsby temperature profile. Even if the CPN-4 radiated plenty of power to this altitude in the beam's 45° top edge the geometry requires that inbound echoes from this layer would disappear at 8.3 miles, not appear for the first time at 8 miles, and would remain invisible for about 7 minutes. In fact power levels will be dropping rapidly at the edge of the beam (conventionally this is the half-power or 3 db-down point) and effective range will be falling back highly nonlinearly, so 7 minutes of signal loss is a very conservative minumum for atmospheric structures that might a priori be expected to have very much poorer direct power reflection coefficients than the aircraft for which the radar is designed. It is very doubtful if this situation satisifies the description of the track in IR-1-56:
This group was picked up approximately 8 miles southwest of RAF Station Bentwaters and were tracked on the radar scope clearly until the objects were approximately 14 miles northeast of Bentwaters. At the latter point on the course of these objects they faded considerably on the radar scope.
Operators are familiar with the way targets move in and out of their radar coverage at different altitudes and are used to making inferences on this basis in the absence of accurate heightfinding. An unusual length of signal loss that indicated targets at 30,000' would have led the operators to suggest an aerial search for the targets at high altitude. But when such a search was made the aircraft were directed to the area at between 2000' and 5000', as discussed below.
Direct backscatter from a structure at a much lower level of a few thousand feet might overcome these difficulties. But there is no clear aerological evidence (outwith the report itself) for any strongly-developed refractive discontinuity at low levels, and the wind speeds of 30 knots or less recorded at these levels are too low to generate 105-knot mean echo speeds even for partial forward scattering to the ground, as has been shown, and for direct reflection the mean speed of the wave or scattering region must be equal to the mean displayed speed on the scope.
Propagation specialist G. D. Thayer expressed broad agreement with an earlier version of the analysis given in Part 2, but remained of the opinion that several points strongly suggested AP. (I should emphasise that the detailed Met. Office records obtained by Paul Fuller were not available at that time.) He pointed to the apparent uncertainty about the number of targets ("12 to 15"), suggesting possible intermittency from scan to scan. He was content with the general correspondence between approximate reported speed and heading and the winds aloft over the major part of the track (notwithstanding the reported motion off-scope at 80° to the wind). And he believed that the negative visual search of the apparent target area by a T-33 was a strong argument for a propagation anomaly. He stated:
I do not, however, believe that these echoes could have been produced by any such mechanism as partial reflections from elevated layers, for precisely the reasons you detailed so carefully . . . No, echoes such as these are apparently produced by regions of clear air turbulence [CAT]. We have much to learn about this mechanism. David Atlas has probably done more work on this problem than any other researcher in the field . . . . Atlas has observed variations on the order hundreds and even thousands of N units [parts per million of refractive index] over distances of only a few meters under such conditions.
Thayer argued that the target behaviour is suggestive of forward scatter from localised domains of CAT induced by wind shear across an inversion boundary which could return ground-incident energy more efficiently than a rippled inversion alone. As he later wrote:
There are two kinds of partial reflection: specular and incoherent. Specular reflections . . . are usually caused by layers of sharply-different refractive index that normally lie more or less horizontally in the atmosphere. These would be unlikely, in and of themselves, to have caused anything like the behaviour of the Bentwaters group echoes. Incoherent echoes are usually caused by CAT, as you mention. This type of reflection is more often described as "forward scatter." If produced by CAT (in the normal sense of the term) such scattering would not cause the kind of echoes seen at Bentwaters that night. The likely culprit, if indeed these echoes were caused by an atmospheric process, is a sort of border-line case. This occurs when two layers of air having a sharp difference in temperature and/or humidity are moving with respect to each other. In other words there is a wind shear across the boundary. When this happens small perturbations in the flow can cause wave-like phenomena to "glide" along the interface between the two layers.
These turbulent patches may scatter radar energy forward to ground targets further away, and back again, in a similar manner to a partially reflecting layer but with greater local efficiency. The scattering is in this case due, as with CAT, to the dielectric inhomogeneity of the turbulent air, and typically results in groups of intermittent moving echoes.
That these patterns tend to come and go can easily be overlooked by the radar operator, or even masked by the antenna rotation period. If there are 12 targets on one scan and 14 on the next, perhaps slightly displaced from where they ought to have shown up, the radar operator will usually assume that they represent the same group of objects he saw on the previous sweep. This mechanism is most effective when the wind speeds are (1) fairly low (say, under 40 knots or so), (2) the wind shears are not large, and (3) the wind speeds generally increase with altitude. These are precisely the conditions at Bentwaters that night. These patches, by the way, can and usually do move at a slight angle to the prevailing wind (10 to 20 degrees or so) and at a different speed, either faster or slower than the prevailing wind speed near the boundary (or boundaries - multiple layers of this type are common).
Thayer pointed out that the motion of the scattering zones can be considered analogous to the waves produced by a pebble in a still pond: the speed of the individual waves is greater than the speed of expansion of the wave-packet as a whole, with new waves forming at the trailing edge and travelling out to the leading edge where they disappear. In a similar way the echo-producing patches of quasi-CAT can exceed the prevailing wind speed, and could possibly in this case produce targets moving at displayed speeds consistent with the operator's estimated range of 80-125 mph.
Of course it remains the case that the forward-scattering efficiency of any atmospheric structure, or travelling region of intermittent similar structures, will be inversely proportional to the grazing angle whatever the given absolute efficiency of the mechanism at a given incidence. Qualitatively speaking the behaviour reported (fading with an increase in range) therefore remains to be explained. But it can be supposed that the reported fading represents either 1) a fluctuation in the inherent reflectivity coefficient of the structure(s), of sufficient amplitude to mask a smaller increase due to the improving geometry, or 2) a change in the back-scattering efficiency of the terrain upon which the reflected radar energy is impinging.
I think it is noteworthy here that since the out-and-back ray path due to a grazing reflection from a layer at low altitude will not differ markedly from the true ground range to the back-scattering terrain, the displayed range of the echo will therefore be similar to the true range of the terrain responsible. A noteable attenuation of echo intensity at a displayed range of 14 miles in fact corresponds quite closely to the range at which the forward scattered signal from such a structure moving roughly NE would cross the coast (somewhere near Dunwich, Suffolk) and begin to impinge on the sea. True ground range to the coast on a 45° bearing is some 18 miles; but bearing in mind that the target cluster is spread over several miles on the range axis, and that Whenry reported the "course flown by this group of objects had slight deviations from SW to NE", the match is arguably close enough.
Energy which was efficiently back-scattered by land textures would, given the probable calm-sea conditions (surface winds 5-10 knots), tend instead to be specularly reflected away from the antenna, consistent with the report that the echoes "faded considerably" at just this time. This seems to me to be quite strong internal evidence for the forward scattering hypothesis, although the fact remains that there is no clear evidence in the Hemsby vertical profiles for the sort of extensive and severe low-level refractive index discontinuity implied by the hypothesis.
Several other points remain to be considered. Firstly, the description of the cluster of echoes bears amplifying in view of the indications of order in this pattern.
The 12-15 targets were
preceded by 3 objects which were in a triangular formation with an estimated 1800 feet separating each object in this formation. The other objects were scattered behind the lead formation of 3 at irregular intervals with the whole group simultaneously covering a 6-7 mile area . . . these objects appeared as normal targets on the GCA scope.
On the hypothesis of forward scatter the trailing group can most easily be interpreted as intermittent stochastic echoes, but it seems necessary to assume that a stable "formation" of 3 targets, distinguished in the report from the "12-15" trailing echoes, must relate to dynamically stable features of the layer propagating across its surface: It cannot relate to ground features, as is obvious from the fact that the forward scattered energy would have a moving footprint impinging on different areas of the terrain at roughly 240-yard intervals (for 125 mph and a 15 rpm scan) over a distance of more than 15 miles, then on the surface of the sea for a further 30 miles or so, before the echoes reached the point at which they merged on the display.
Because the reported disposition of this "formation" of 3 echoes is rather specific, the degree of dynamic stability required over many minutes is not altogether easy to entertain. It is possible to assume that the ordered "formation" was only observed momentarily, a chance feature which soon evaporated; but there really is no justification for this in IR-1-56 whose clear intent is to imply that this pattern was a significant feature of the echoes "prior to consolidation into one object".
Secondly the evidence indicates forward scattering associated with a layer at an altitude of no more than a few thousand feet. This is suggested by several independent arguments: a) The rough equivalence of the implied length of the scattered ray path at the point of fading and the true ground range to the coast, and b) the close-range observation of the echoes traversing between opposite scope sectors. Together these require a narrow grazing angle and first-trip returns. c) The assigned search altitude of the T-33 was between 2000 and 5000'. This introduces a difficulty in terms of likely echo speeds.
In Thayer's opinion, "personnel at the radar site believed the targets were at about 3500' or thereabouts. They would have been able to estimate this altitude based on the behaviour of the echoes as they appeared to overfly the radar (the distance when they disappeared and then reappeared, together with the known vertical radiation pattern of the antenna, would be a clue as to the probable altitude of the objects producing the echoes." Given this reasonable assumption, then the vertical pattern of the CPN-4 beam allows us to infer that the beam-angle disappearance of the targets occurred when at a displayed range of about 5000'. Because we have concluded that the displayed range must be the real range (Section d; remember that delayed multiple-trip echoes that could display at spuriously close ranges cannot cross between opposite scope sectors) the reflection geometry of this situation, governed by the known 45° upper edge of the CPN-4's cosec2 coverage, leads to a maximum layer altitude of no more than about 2000'. But interpolating the Bentwaters and Hemsby weather data indicates that winds at 2000' can have been no more than about 20 knots, leading to mean displayed echo speeds due to forward-scatter of only around 40 knots, or less than half the mean measured target speed.
Moreover, even this maximum estimate of 2000' already requires very efficient reflectivity at an abnormally high incidence of about 45°, a figure several times the typical grazing angle above which signals normally fall below the threshold of detectability even for very sensitive radars; and here we are considering a low-power airfield surveillance radar of vintage circa 1956. Therefore one is led to suppose either very remarkable atmospheric conditions indeed at 2000' - for which there is neither general nor particular justification in the meteorological evidence - or a marginal scattering layer considerably lower than 2000' which can be intercepted at a grazing angle at a slant range of 5000'. The geometry of this situation would point to a sharp discontinuity in the N-profile at less than 100 feet. Interestingly this would be consistent with the top of a strongly subrefractive surface layer indicated in the Hemsby profile, suggesting the possibility of dielectric inhomogeneities due to turbulence associated with mixing between the humid surface layer and overlying drier air. According to Blackmer et al.  it has been suggested that such conditions might cause detectable echoes on modest radars. However, with windspeed near the surface of around 10 knots it becomes quite impossible to explain the observed echo speeds by this mechanism.
The question of target speed of course remains uncertain, resting as it does on resolving the discrepancy earlier noted between speed as derived from tracklength-over-time and the cited operator estimate. How confident ought one to feel about the speeds reported?
It was suggested that the range of speeds quoted may have been quoted out of context by the preparing officer and could in fact refer to only a limited portion of the track. Perhaps it was offered as a rough average. In this case, if the reported total duration and the durations of the stationary episodes were accurate, then the spread of implied values would be extreme - between 0 and several hundreds of mph - which would of course be grossly inconsistent with the forward scattering model. Alternatively one can plug in different assumptions to square the overall duration of the track with the reported range of speeds.
For example, it may be significant that by eliminating the reported stationary periods the cited velocities and the 2130-2155 duration can be nearly reconciled, since at 125 mph the echoes would travel about 52 miles in 25 minutes, bringing them roughly to the point of merger (at 80 mph track length would be 33 miles, and at the mean of these speeds about 43 miles). The top reported speed is still inconsistent with a required mean of 170 mph, but arguably not dramatically so if one is prepared also to to ignore ad hoc the subsequent northward motion of the integrated target cluster off-scope, an expedient which happily would further improve the fit to the forward-scattering model.
Let us suppose that a transcriptive error in the reporting chain has resulted in the stationary episodes being given in minutes instead of seconds. Such an error seems possible, whereas it seems highly improbable that so singular a statement as that the targets ceased all motion twice could be imported in error. In this way the substance of the report is preserved but the stationary episodes would be brief enough to keep the implied speed (tracklength-over-time) closer to the operator-measured speed within the limits of acceptable error.
Unfortunately an observed stationary period of 3-5 seconds (instead of the "3 to 5 minutes" reported) would make no sense in terms of the scan rate of the CPN4 and the very small scan-to-scan spot displacement on the PPI. This period would correspond to only a single 4-second scan and if any such brief stop could be observed it would be reported by the operator as 4 seconds, or a multiple of this period, but never 3 or 5. Furthermore, at a displayed speed of 170 mph one sweep corresponds to a range increment of about 1000' or 0.32% of the display radius (assuming that maximum range scale is selected, as would be the case for a target being observed outbound at >40 miles range), and since between 150 and 200 spots can typically be resolved along the radius of a magnetically deflected analogue display of this type [Haworth 1948] corresponding to a resolution of about 0.5% of the tube radius, a radially moving 170 mph target would need at least two scans to be displaced more than its own spot diameter on the PPI - even if it were not an abnormally large echo.
Indeed, the range increment of 1000' is only twice the smallest possible range resolution of a 1-microsecond pulse and would only barely be present between signals fed to the display. A pause of some 4 seconds, therefore, would be indiscernable in principle. Even a pause of 10-15 seconds (instead of the 10-15 minutes reported on the other occasion) would occupy no more than 3 or 4 scans, and the displacement of the moving target over 4 scans would be only about double the smallest resolution of which the cathode ray tube is capable. Again, bearing in mind that we are considering a very large echo presentation at this point, no such pause is likely to have been observed, let alone thought worthy of particular note by the operator.
Given that this inelegant attempt to discount the stationary episodes, even if successful, would still have left an awkward internal inconsistency in the estimated speeds, it was clearly never going to be ideal. There appears to be no good reason to discount the report that the target(s) maintained station for between 13 and 20 minutes which, if the overall duration is correctly given, means that the true upper target speed must have been very much higher than the speeds said to have been computed by Sgt. Whenry, with fatal consequences for the forward scatter hypothesis.
The crucial clause here, of course, is: "if the overall duration is correctly given." A close reading of IR-1-56 suggests that the simplest interpretation of all - that Captain Holt or one of his clerical personnel mis-transcribed the time - may well be the correct one. If so, this shows how the merest of mistakes can create all kinds of confusion, and underscores the risk - at a distance of 40 years - of accepting even official records without careful interpretation. By the same token, it is chastening to bear in mind that, once corrected, this mistake reveals Whenry's UFO report to be fully self-consistent.
IR-1-56 states that the Control Tower Shift Chief, Sgt. Wright, began observing the amber light in the SE sky (which we can be confident was the planet Mars, azimuth 116°, elevation 9° at 2200Z August 13) at 2120Z:
Sgt. Wright indicated that his attention was first called to the object by its position, size and unusual colour. He was also aware that the Bentwaters GCA was tracking Unidentified Flying Objects at this time.
But according to the radar reports the earliest acquisition of any of the Bentwaters targets was at 2130Z, the time when according to Captain Holt the slow cluster of targets on Track D and the rapid target on Track A were both first detected. That is, Sgt. Wright cannot have been aware that GCA was tracking targets "at this time" unless the stated time of radar acquisition is wrong; or conversely, Sgt. Wright cannot have begun his visual observation until at least ten minutes later than stated. Obviously one of these times is wrong, and since there is already a suspicious discrepancy in the times attributed to Whenry it would be reasonable to suppose that Track D did not commence at 2130 but somewhat earlier, thus explaining Sgt. Wright's otherwise mysterious premonition.
Now the simplest hypothesis here is that a single digit in the group "2130Z", which appears just once in the report of Whenry's tracking, has been mistyped - probably by confusion with the 2130 time of Vaccare's Track A. Given the reported end time of 2155 and the context of other events the error cannot be in either the "2" or the "1", and the true time must be significantly before 2120 since Sgt. Wright, in the Control Tower, was already aware by this time that GCA personnel, in their radar trailer out on the airfield, had been tracking UFOs. Therefore, sticking to the well-worn principle of Occam's Razor that the simplest hypothesis is to be preferred and continuing to assume that only a single digit has been mistyped, one is led to conclude that the offending digit must be the "3" and that the true time can thus only have been either 2110 or 2100. And this latter value rather neatly resolves the speed problem: Deducting 15 minutes to allow for the stationary episodes from a total duration of 55 minutes yields a mean overall speed required of <110 mph, close enough to the mean (102.5 mph) of Whenry's bracketed estimate to give us confidence that this hypothesis is very probably correct.
The clincher is the statement elsewhere in IR-1-56 that Lts. Metz and Rowe in the T-33 began their aerial search at 2130Z, and "at this time" were vectored "to the north east of Bentwaters to search for unidentified flying objects that were being tracked by Bentwaters GCA." In other words, by 2130 the targets had already crossed the scope centre from the SW into the NE sector and GCA personnel had had the time to consider and initiate action when the T-33 adventitiously arrived in the area. This means that the radar tracking of these targets in the SW quadrant simply must have begun quite some time before 2130, thus allowing the Control Tower to have become aware of it by the time of Sgt. Wright's observation at 2120.
(Note: Further support for this hypothesis comes from the recent discovery that at 2120Z an RAF Venom jet crewed by Flying Officers Arthur and Scofield was scrambled from RAF Waterbeach and sent NE to intercept an unidentified radar target. Some few minutes into their flight, very possibly at about 2130Z, they lost their wing tanks and had to abort. This would coincide neatly with the decision of Bentwaters to request assistance from the nearby T-33 "at this time". See here.)
Some of the problems of this report can therefore be resolved in favour of the forward-scatter hypothesis, notably the range of fading of the targets which is rather convincingly related to the range at which the radar footprint would begin to impinge on the specularly reflecting surface of a calm sea. The restored duration permits echo-speeds of a conceivable order for meteorological structures. And the hypothesis is also circumstantially supported by the failure of the T-33 search to find any aerial objects in the area (although it should be said that this last point is not entirely unambiguous given the lack of airborne radar, the dark-sky conditions, and an uncertain target altitude - tracking out of the maximum drum on the PPI, for example, would not exclude the possibility that the targets climbed above the inferred search altitude during recession to the NE), .
In summary the forward-scatter model, or something like it, is on the whole the least unattractive explanation of the reported echoes. But it is reasonable to point out some difficulties thatremain:
|a) The reported off-scope motion
of the merged target, which was on a changed "north"
heading fully 80° away from the wind for 20 miles. Moreover
this 80° 'error' is in a direction opposite to the
mesoscale cyclonic circulation of the air mass over the
UK at the time (see Part 2 and Meteorological
Data). It is
possible to suppose reportorial error here, but this
would be ad hoc with no internal justification.
b) A prior mean target heading 35° away from the most favourable winds aloft. The mismatch between target heading and winds is therefore probably somewhat larger, and somewhat more consistent, than might normally be expected for patches of turbulence moving for 55 minutes under the influence of winds. This alone is not necessarily a strong objection, however.
c) Indications that a forward scattering layer would, ex hypothesi , have been at a low altitude (order of 2000' or less) subject to light winds (order of perhaps 10-20 knots) very unlikely to yield a mean echo speed of 100 mph and a measured maximum of 125 mph.
d) The long duration of the two stationary episodes. The mean estimated duration of the first episode would occupy some 190 scans of the antenna; the second some 60 scans. Note that whilst Whenry's estimate of elapsed time is in each case an approximation, these mean durations plug in well to the model we now have of the track derived from speeds measured, distance covered and the restored 55-minute overall duration. Thus the reported absence of significant movement within the practical limits of PPI resolution for dozens of antenna revolutions has to be taken seriously. That these stationary episodes punctuated periods of steady linear movement under the influence (ex hypothesi) of winds seems in need of some explanation.
e) The implied dynamical stability of detailed features of the forward-scattering layer. Here we have a fairly compact group of targets in linear motion over a period of several tens of minutes and over a distance of several tens of miles. At Washington in 1952, when balloon soundings showed the presence of a distinct low level temperature inversion, clear air targets observed over a comparable period by Borden & Vickers  on the S-band surveillance radar were distributed stochastically in location and time. They were noteably weak and fluctuating (see g. below) and the average distance over which it was possible to track an echo continuously was only 2.1 miles
f) The appearance of order in the disposition of the echoes: in particular the "triangular formation" of 3 targets which was noted as distinct from the "irregular" scatter of the "12 to 15" targets in the trailing cluster.
g) Presentation comparable to "normal targets", noteably when at close displayed slant range (high elevation) despite the unfavourable centimetric frequency, modest peak power and poor sensitivity of a short-range airfield surveillance radar not optimal for detecting the dielectric inhomogeneity of clear air structures.
Comparable experience with such structures at S-band on airfield surveillance and ATC radars at Washington and Indianapolis in 1952, in conditions where temperature inversions were measured, showed that such targets were characteristically poor and "usually easy to recognise" as different from normal aircraft targets [Borden & Vickers, 1953]. At Washington in August 1952 with a surface inversion in the first few hundred feet the echoes were "uniformly small and usually had a weak, fuzzy appearance. However the target intensity varied from sweep to sweep. Occasionally one or two strong returns would be received in succession, followed by almost total blanking". At Indianapolis in November with an inversion at 6000' the targets were "larger, stronger and more numerous" on the PPI. They were studied closely with an A-scope which showed slow rise and decay times and wide variations in amplitude distinct from the sharp and steady profiles of aircraft targets. The signal envelope of the clear air echoes was similar to that returned from rain clouds.
h) Signal attenuation not inversely proportional to displayed range (or not proportional to elevation), as is characteristic of partial forward-scattering.
i) No direct aerological evidence of refractive index stratification at levels consistent with the hypothesis. The layer thickness for clear air scattering would be less than the wavelength - in this case < 10 cm - and it is plausible that because such shallow layers will tend to fall between the balloon samples they will almost never be measured directly. On the other hand, the canonical Borden & Vickers model of clear-air scattering developed from observations at Washington and Indianapolis in 1952 was based on the fact that although the structures responsible could not be directly observed nevertheless in almost every case the onset of the 'angel' echoes was correlated with balloon measurements of a temperature inversion in the lower troposphere. The general meteorological conditions were classically 'hot and muggy' and the half-speed of the rather ephemeral echoes in every case correlated with the wind speed at an appropriate inversion level. In the present case echoes of unusual discreteness, stability and persistence are implied although the general conditions are more ambiguous, and there is no radiosonde evidence of inversion at any appropriate level.
* * *
In conclusion the only realistic hypothesis which comes close to explaining Track D is that of highly efficient forward scatter and/or direct backscatter from turbulent clear-air features or waves propagating across a sharp temperature/humidity boundary layer. There remain some problems with this interpretation, however. It may be safe to say that the targets suggest the presence of some sort of atmospheric radio-scattering structure with properties on, or just beyond, the margins of current knowledge. It is not clear, of course, that this is necessarily different from calling them "UFOs".
© Martin Shough