Earhart Project Research Bulletin
|by Bob Brandenburg, M.S.
18 June 2007
BackgroundAn unpublished 1996 paper by Charles N. Hill, entitled “BRIDGE TO THE MARSHALLS, The Earhart Flight to the Rising Sun,” excerpts of which were published in 1998,1 has come to the attention of this author,2 and requires comment.
Hill concludes that Amelia Earhart ditched her Electra near the Marshall Islands in July 1937, and sent radio signals from her floating plane for several days afterward. His conclusion is based on two premises: (1) Earhart’s radio transmitter could operate if her plane was afloat; and (2) all bearings on radio signals from the central Pacific – southwest – taken by the Pan American Airways (PAA) radio direction finder (RDF) at Mokapu Point, on Oahu, were in error by 35.2° because the system was not calibrated in the southwest sector.
Hill’s conclusion could be true only if both his premises were true. The analysis in this paper shows that both premises are false, and therefore Hill’s conclusion is false.
Time zones used herein are PST (Pacific Standard Time), HST (Honolulu Standard Time) and GMT (Greenwich Mean Time). HST was 10.5 hours behind GMT. A combined time/date notation is used for brevity, in which the date follows the time zone, separated by a slash. For example, 1200 HST/5 means 1200 Honolulu Standard Time on the 5th day of the month; July is understood as the month for purposes of this analysis. Where GMT is referenced, the “Z” zone time designator is used. For example 1400Z/4 means 1400 GMT on 4 July.
Radio messages cited in this analysis are in the TIGHAR data base.3 All bearings are great circle, and all distances are in nautical miles (nmi).
Could Earhart’s Transmitter Operate if Her Plane Was Afloat?
… RADIO TECHNICIANS FAMILIAR WITH RADIO EQUIPMENT ON PLANE ALL STATE THAT PLANE RADIO COULD NOT FUNCTION NOW IF PLANE IN WATER AND ONLY IF PLANE WAS ON LAND AND ABLE TO OPERATE RIGHT MOTOR FOR POWER.
Hill dismisses this advice:
This came as quite a surprise to all who had heard the Electra during the previous three days, and, considering the lack of consideration of the two Exide 75 Amp-Hour batteries aboard, it apparently originated with those who had had no recent technical contact with the aircraft.
Hill states his own belief that “the Electra’s transmitting capability had been reduced and altered – but not eliminated – by the fact that her transmitter was being powered by a submerged motor-generator set,” and disdains “those who choose to continue to adhere to the myth that Earhart’s Electra could not transmit on the water.”
Hill ignores a COMFRANDIV message to Itasca and COMHAWSEC 6 at 2245 HST/5, stating in pertinent part:
DYNAMOTORS ALL MOUNTED UNDER FUSELAGE AND WOULD POSITIVELY BE SUBMERGED IF PLANE WAS ON WATER.
Hill also ignores a COMFRANDIV message to Itasca at 2330 PST/5:
INFORMATION JUST RECEIVED FROM LOCKHEED AIRCRAFT COMPANY STATES POSITIVELY EARHART PLANES RADIO TRANSMITTER COULD NOT REPEAT NOT OPERATE IF PLANE WAS ON WATER.
Nothing could be clearer than this definitive statement by the Electra’s designer.
It was impossible for the transmitter to operate if the plane was floating, and Hill’s premise is false. It is interesting to note that Hill’s asserted certainty that signals from Earhart were heard “during the previous three days” – when considered in light of the fact that her transmitter could operate only if the plane was on land – is consistent with the TIGHAR hypothesis, which holds that Earhart landed on one of the Phoenix Islands, and transmitted radio signals from there.
Was There a 35.2° Bearing Error in the Mokapu RDF System?
Analysis of the RDF bearings requires at least basic knowledge of the PAA RDF system, skywave radio signal propagation, and the factors affecting terrain interference with signals arriving at the Mokapu site. Hill’s analysis apparently was not informed by the minimum requisite knowledge. A brief summary is provided here as background for the analysis in this section.
The PAA RDF System. The system12 accurately measured a signal’s direction of arrival (DOA), regardless of direction, by virtue of the Adcock antenna array geometry13 and the unique goniometer14 connected to the array. Hence the system neither required nor included means for explicit calibration. However, external factors could skew a signal’s DOA at the RDF site, in which case the system would accurately measure the skewed DOA, and the result would be a bearing error. It is important to note that such bearing errors could not be constant unless the DOA-skewing factors were constant.
The system’s sensitivity pattern was the same as that of a loop antenna, and rotating the goniometer produced the same effect as rotating a loop. The bearing procedure15 was to: (1) rotate the goniometer to get an aural null on a connected radio receiver, then offset the goniometer to one side of the null to get a signal level convenient for measurement, and note the bearing; (2) offset the goniometer to the other side of the null and note the bearing where the same signal level was found; and (3) average the offset bearings – “splits” – to get the signal bearing. Poor reception conditions could cause some uncertainty in the splits, which would not constitute a bearing error, but merely a known degree of uncertainty as to the actual bearing.
Skywave Radio Signal Propagation. The signals heard at the PAA RDF stations were skywaves, which propagate on great circle paths via refraction in the ionosphere. An arriving skywave has an elevation angle determined by the ionization density profile, which is proportional to the sunspot number and varies with season, date, and time. Various propagation anomalies can occur in the ionosphere. A common example is “swinging” or “shifting”, in which the observed DOA oscillates around an average value, instances of which are mentioned in the PAA RDF reports.16
Terrain Interference. The Koolau Range17 extends from the northern tip of Oahu, near Kahuku Point, to the southeastern tip at Makapuu Point. The ridge line orientation is generally northwest to southeast, with significant localized deviations. The southwestern slope18 is a complex forested watershed with ridges, valleys, and running streams. The RDF site19 was on Mokapu Point, about 12 nmi northeast of Honolulu, and about 6 nmi northeast of the ridge line. Whether, and to what extent, the DOA of a skywave from the southwest was skewed depended on the ridge line terrain on the source bearing, and the magnitude of the skywave elevation angle (SEA) relative to the ridge line elevation angle (RLEA) – the vertical angle of the ridge line from the RDF site.
The potential for DOA skewing was greatest if the ridge line sloped significantly across the source bearing, and was oriented at an obtuse angle to the source bearing. In this case, a wavefront with an SEA near the RLEA would encounter uneven terrain “drag,” resulting in DOA skewing analogous to that caused by the well-known coastal refraction effect in the case of a ground wave. Signals with larger SEAs would be subject to less skewing, and signals with SEAs several times the RLEA would experience little or no DOA skewing.
In contrast, there was no potential for DOA skewing if the ridge line did not slope across the source bearing, and was perpendicular to the source bearing. In this case, the average DOA of a signal with SEA greater than the RLEA would be a good estimate of the source bearing.
In any case, regardless of DOA skewing, a low-SEA signal could be attenuated by ground absorption loss.
Hill’s premise is based on a single bearing taken at Mokapu during the evening of 5 July, in response to Itasca’s 2145 HST/5 request to COMHAWSEC:
FOR PURPOSE DETERMINING CORRECTIVE FACTOR ON EARHART BEARING REQUEST HONOLULU TAKE BEARING ON ITASCA ON 3105 KCS AT 2215 AND 2315 HONOLULU TIME ALSO WAKE AT 2315 HONOLULU TIME.“Honolulu” is a reference to the Mokapu RDF site.
Itasca transmitted20 at 2213 HST/5, and from 2315 to 2318 HST/5. A severe electrical storm near Wake Island21 prevented signal reception there. COMHAWSEC told Itasca, at 0138 HST/6, that Mokapu got a bearing of approximately 196.5° at 2317 HST/5, “BELIEVED TO BE ON ITASCA PHONE TRANSMITTER.”
Hill assumes that Mokapu readily identified Itasca:
That Mokapu had no problem identifying Itasca for the corrective bearing, Mokapu 196.5, is confirmed by the COMHAWSEC log transcripts, … at about the very times scheduled by Itasca for the corrective bearings, the Itasca was clearly heard and identified by ComHawSec on 3105 KHZ …
Hill’s reasoning is faulty. COMHAWSEC identification of Itasca does not confirm identification by Mokapu. The reception conditions were different at the two locations. COMHAWSEC’s radio facilities22 were in the Honolulu Aloha Tower, with no terrain to the southwest. But Koolau terrain interference could have attenuated the signal at Mokapu, preventing identification there.
The corrective factor to be applied to all of the Mokapu bearings, then, is 35.2 degrees. That is, 35.2 degrees must be added to all three of the Mokapu RDF bearings, to correct them for application in the Earhart search. This extreme RDF error is understandable when one considers that Mokapu had no reason to be calibrated to cut bearings in the area of Itasca or even the “eastern edge of the Marshalls.”
Hill’s corrective factor is the difference between 196.5° and his estimate of Itasca’s bearing, 231.7°, at the time. Itasca’s actual bearing23 was 231.9°.
As will be shown below, Hill commits the fallacy of Hasty Generalization by basing his premise on a single bearing observation.
The three Mokapu RDF bearings are the 196.5° test bearing, taken at 2317 HST/5, and two bearings believed to be on Earhart signals. The ionosphere ionization density profile at 2317 HST/5 was determined by solar radiation during daylight on 5 July, when the sunspot number (SSN) was 91.24 The first Earhart bearing was 213°, just before sunrise on 4 July. The ionization density profile at that time was determined by solar radiation during daylight on 3 July, when the SSN was 74. The second Earhart bearing was 215°, at an unspecified time during the night of 4/5 July. The ionization density profile that night was determined by solar radiation during daylight on 4 July, when the SSN was 65. Clearly, the ionization density profile was different at the times of the three bearings.
As shown earlier, PAA RDF bearing errors were caused by DOA skewing, not by lack of calibration. Since DOA skewing of signals arriving at Mokapu from the southwest was a function of SEA and the Koolau ridge line terrain on the source bearing, and since the three Mokapu bearings were taken on different sources at different times on different dates, Hill’s assertion that “35.2 degrees must be added to all three Mokapu RDF bearings” is tantamount to a conjecture that both the SEA and the ridge line terrain on the source bearing would be the same as in the case of the 2317 HST/5 bearing, for any signal arriving at Mokapu from anywhere in the southwest, at any time on any date.
Disproving Hill’s Conjecture. Any counterexample disproves Hill’s conjecture. Since Hill’s conjecture applies to all potential signals from anywhere in the southwest, at any time on any date, it applies to all potential signals from anywhere on Itasca’s track25 during the night of 5/6 July. Consideration of potential signals from Itasca during that night reveals three counterexamples. Bearings and SEA values in the following analysis were computed with the ICEPAC26 model.
Itasca’s bearing from Mokapu stayed within a 3.1° sector during the night. Itasca’s 1930 HST/5 position was 5° 08´ north latitude, 177° 43´ west longitude, bearing 232.7°. The 0530 HST/6 position was 5° 15´ north, 175° 40´ west, bearing 229.6°. The terrain configuration27 is virtually the same for all bearings in that sector. Hence SEA was the only variable DOA skewing factor, and a deviation from the 2317 HST/5 SEA value is a counterexample to Hill’s conjecture.
The RLEA in Itasca’s bearing sector is 4.7°. The ridge line28 there slopes down to the right, across the wavefront direction of travel, at an angle of 9.6°, and crosses Itasca’s bearing at an angle of 120°, i.e. 30° counterclockwise from perpendicular. The 2317 HST/5 SEA was 5.9°, consistent with terrain interference causing signal attenuation due to ground absorption loss, as well as DOA skewing due to uneven “drag” by the sloping terrain.
With three exceptions, the SEAs for potential signals at hourly intervals along Itasca’s track were between 3.2° and 5.2°, suggesting that the DOA skewing in those cases was about the same as at 2317 HST/5. The exception cases were:
In each exception case, the SEA was about 3 times the 2317 HST/5 value, clearly a significant deviation and thus a counterexample. Therefore, Hill’s conjecture is disproved.
The bearing error observed at 2317 HST/5 was a random observation of a time-dependent variable, resulting from Itasca’s choice of test time, and not from some supposed constant condition(s) in the southwest sector. There was no constant 35.2° error in the Mokapu bearings.
The Other Two Mokapu Bearings.It is instructive to examine the circumstances of the other two Mokapu bearings on signals that Hill considered as being from Earhart, and to which he erroneously applied a 35.2° “corrective factor.” As stated earlier, the first of these bearings was 213°, and the second bearing was 215°. Both bearings pass within about 30 nmi of Gardner Island, now known as Nikumaroro (Niku), in the Phoenix Islands. Niku is at latitude 4° 40´ south and longitude 174° 32´ west, approximately 1850 nmi from Mokapu, bearing 214°, and is the island where the TIGHAR hypothesis holds that Earhart landed her Electra.
It is significant to note that on bearings near 214°
In light of this terrain configuration, the potential for DOA skewing of signals from bearings near 214° was virtually nonexistent at any SEA, and there is no basis for applying any correction to either bearing.
The Mokapu report29 states this signal was heard from 1523Z/4 to 1530Z/4 (0453 HST/4 to 0500 HST/4) and describes it as:
CARRIER AGAIN HEARD ON 3105 – ROUGH BEARING ONLY POSSIBLE DUE TO WEAKNESS AND SWINGING OF SIGNALS. GET BEARING FROM MOKAPU OF APPROXIMATELY 213 DEGREES. ADVISED COAST GUARD.
COMHAWSEC relayed the report to Itasca at 0540 HST/4:
FOLLOWING FROM MOKAPU PAA STATION AT 0445 TO 0500 ROUGH WEAK SIGNALS SPLITS BADLY DOUBTFUL BEARING OF 213 ON GONIO MINIMA MAY BE PLUS OR MINUS TEN DEGREES BAD SHIFTING OFFERED ONLY AS A POSSIBILITY.
The difference between the reported signal starting times is not critical.
The phrase “SPLITS BADLY DOUBTFUL BEARING OF 213 ON GONIO MINIMA MAY BE PLUS OR MINUS TEN DEGREES BAD SHIFTING” means the operator could not get reliable offset bearings using the standard procedure, so he took the bearing directly in the aural null and estimated that there could be as much as a ten degree error because the signal bearing was shifting (swinging), an anomalous ionospheric propagation mode discussed earlier in this paper. Hence, 213° is the average estimated bearing.
The signal was heard at Mokapu during a period when the SEA was between 9° and 13°, approximately 3 to 4 times the RLEA, suggesting minimal ground absorption loss. The reported signal weakness is consistent with ionospheric absorption loss. But despite the weakness and shifting of the signal, the absence of DOA skewing means that 213° was a reasonably good estimate of the source bearing.
Since Earhart’s transmitter could operate only if on land, the evidence suggests that the signal originated at Niku.
The Mokapu report30 shows that this bearing was taken at an unspecified time between 2000 HST/4 and 0155 HST/5, during the second evening of broadcasts by Honolulu radio station KGMB requesting Earhart to respond, and describes the signal as:
CARRIER HEARD FROM DIRECTION FINDER CLOSE TO 3105 BUT SIGNALS SO WEAK THAT IT WAS IMPOSSIBLE TO OBTAIN EVEN A FAIR CHECK. AVERAGE SEEMS TO BE AROUND 215 DEGREES – VERY DOUBTFUL BEARING.
COMHAWSEC did not relay this bearing report to Itasca, so the actual bearing time is unknown. But whenever the bearing was taken, a signal from a source bearing approximately 215° would arrive in the same ridge line area as in the 213° bearing case discussed above, and DOA skewing would not be a factor. Despite the weakness of the signal, the absence of DOA skewing means that average bearing was a reasonable approximation of the signal source bearing.
As in the case of the 213° bearing, since Earhart’s transmitter could operate only if on land, the evidence suggests that the signal originated at Niku.
Hill’s conclusion – that Amelia Earhart ditched her Electra near the Marshall Islands in July 1937, and sent radio signals from her floating plane for several days afterward – is based on false premises, and therefore is false.
Hill’s premise – that Earhart’s radio transmitter could operate if her plane was afloat – shows ignorance of the Electra and basic principles of electrical equipment. Hill also disregarded a definitive statement from the Lockheed Aircraft Company – the Electra’s designer – that the transmitter could not operate if the plane was on water. Careful examination of the relevant factors shows that Hill’s premise is false. Therefore, any radio signals heard from Earhart after her disappearance were transmitted from land.
Hill committed the fallacy of Hasty Generalization in relying on a single test bearing on the Coast Guard cutter Itasca as the basis of his premise that all three bearings on signals from the southwest taken by the PAA RDF system at Mokapu Point were in error by 35.2°. Hill’s analysis in support of his premise apparently was not informed by requisite knowledge of the PAA RDF system, skywave radio signal propagation, or potential terrain interference by the Koolau Range. The observed test bearing error was a random observation of a time-dependent variable, resulting from Itasca’s choice of test time, and not from some supposed constant condition(s) in the southwest sector. Hill’s premise was easily disproved by counterexamples. There was no constant bearing error.
There was no basis for applying a “correction” to the other two Mokapu bearings, the PAA-reported estimates of which passed within about 30 nmi of Nikumaroro, and which were erroneously cited by Hill as “proving,” when “corrected” by 35.2°, that Earhart ditched near the Marshall Islands. The bearings as reported by the PAA RDF site were reasonable approximations of the true source bearing, and the evidence suggests that both signals originated at Nikumaroro.
|1||Prymak, Bill, “Follow-up On The PAA Post-Flight Radio Intercepts,” Amelia Earhart Society Newsletter, March 1998. BACK|
|2||Bright, Ron, personal communication to the author, 20 March 2007. BACK|
|3||Available on the DVD in Finding Amelia, by Ric Gillespie, published by U.S. Naval Institute Press, 2006. Finding Amelia. BACK|
|4||Commander of the San Francisco Coast Guard Division. BACK|
|5||Coast Guard cutter searching for Earhart. BACK|
|6||Commander of the Coast Guard Hawaiian Sector. BACK|
|7||Lockheed Electra 10 Service Manual, available from TIGHAR. BACK|
|8||Morgan, Howard K., Aircraft Radio and Electrical Equipment, Pitman Publishing Corporation, New York and Chicago, 2nd Edition, 1941. BACK|
|9||Harney, Bill, detailed diagram based on original Lockheed drawings, in TIGHAR archive. BACK|
|10||Gillespie, Ric, personal communication to author. BACK|
|12||U.S. patent 2174014, www.uspto.gov. BACK|
|13||U.S. patent 2166100, www.uspto.gov. BACK|
|14||U.S. patent 2174017, www.uspto.gov. BACK|
|15||Sandretto, Peter C. Principles of Aeronautical Radio Engineering, (New York and London, McGraw-Hill, 1941), p. 254. BACK|
|16||“The Pan American Airways Memos,” Research Document 34, TIGHAR website. BACK|
|17||U.S. Geological Survey map NF 4-11, Oahu. BACK|
|18||U.S. Geological Survey maps, “Mokapu Quadrangle,” “Kaneohe Quadrangle,” and “Honolulu Quadrangle,”,7.5 minute series, scale 1:24,000. BACK|
|19||Brandenburg, Bob, “Analysis of Radio Direction Finder Bearings in the Search for Amelia Earhart,” TIGHAR website. BACK|
|20||See message data base on DVD, op cit. BACK|
|21||“The Pan American Airways Memos,” Research Document 34, TIGHAR website. BACK|
|22||U.S. Coast Guard Communications Station Honolulu unit history, www.uscg.mil/d14/units. BACK|
|23||Based on Itasca’s navigation data, on DVD in Finding Amelia, op cit. Finding Amelia. BACK|
|25||See navigational data base on the DVD in Finding Amelia op cit. Finding Amelia. BACK|
|26||Developed by the U.S. Department of Commerce Institute for Telecommunication Sciences, Boulder, CO. Available at www.its.bldrdoc.gov. BACK|
|27||U.S. Geological Survey map “Kaneohe Quadrangle,” 7.5 minute series, scale 1:24,000. BACK|
|28||U.S. Geological Survey map “Kaneohe Quadrangle,” op cit. BACK|
|29||“The Pan American Airways Memos,” Research Document 34, TIGHAR website. BACK|