The entry says that the second question Castle asked, where the bomb’s debris would go, was answered by a drawing, and that the drawing left out the top of the cloud. This piece is about the drawing: what it was made from, how it was read on the night, what the wind did and did not do, and why, sixty years later, the official reviewers of the test concluded that neither the yield nor a shift of the winds was the cause of what fell on Rongelap.

The five per cent

The planning for Castle’s radiological safety was, in the Defense Threat Reduction Agency’s words, “largely a continuation of that done for the two high-yield shots detonated at Eniwetok during Operation IVY in the fall of 1952.” The first of those, Mike, had been a surface burst of 10.4 megatons, and almost nothing of it had been found. “Only trace amounts of radioactive fallout had been detected outside of the Danger Area surrounding Eniwetok Atoll. Indeed, most of the radioactive debris from the 10.4 [megaton] surface burst MIKE shot seemed, quite literally, to have disappeared from the face of the Earth, apparently having been trapped in the stratosphere.” Clarkson and Graves wrote after Bravo that “Although conscientious efforts were made to document the fallout from MIKE, only about 5 percent of the total debris could ever be accounted for.”

The explanation to hand was stratospheric trapping. The troposphere, where weather happens, is well mixed by convection; above it, in the stratosphere, temperature rises with height and the air is stratified and quiet, so that what is put there stays there. The dust of Krakatoa in 1883 had sat near 75,000 feet for months and gone twice round the world. If Mike’s debris had gone through the tropopause, it would come down slowly, dispersed and decayed, as “global fallout”, a concern of its own for the strontium it would put in children’s bones but not a hazard to anyone a hundred miles from the shot.

The task force’s staff weather officer, Commander Elbert Pate, took the theory as far as it would go. In a memorandum of 3 March 1953 he wrote that “all evidence indicates that [the] MIKE cloud stem penetrated to a final altitude of about 125,000 ft and that by far the major portion of atomic debris was carried well into the stratosphere”, and recommended that the requirement for a southerly wind aloft, to carry a cloud away from the inhabited islands, be “de-emphasized”: “With extremely high yields, most of the radioactive debris is carried into the stratosphere, where it is ‘trapped’ and falls out very slowly, thus undergoing normal processes of decay, mixing and dispersal. All past sampling operations bear out this conclusion, as do the results of the world-wide fall-out program.” The task force’s operations chief “heartily” concurred, and wrote to Graves that “if we can adopt [Commander Pate’s] concept, the importance of wind direction will become less critical.”

The task force did not adopt Pate’s concept. Its radiological safety report says that “confidence could not reasonably be placed in significant trapping of debris by the tropopause unless definite proof of such a mechanism was available”, and the wind requirement stood. But the drawing it made was shaped by the theory all the same. The initial distribution of debris for Bravo “was taken to be a cylinder 15 mi in diameter extending upward from the shot site to the base of the stratosphere. How much of the radioactive debris would be below the tropopause and how much would remain trapped in the stratosphere was, however, unknown.” Fallout from the head of the cloud, above the tropopause, was reckoned to take more than twelve hours to reach the ground at the standard settling rate of 5,000 feet an hour, and to be dispersed and cooled by then: “Before the BRAVO detonation, the possibility of stratospheric fallout was thus viewed as more of an academic than a safety problem.”

The plot itself was made in two forms. The RADEX, the radiological exclusion area, was the region the task force’s own units were forbidden to enter without authority; the Fallout Plot was the wider area of “significant” fallout, down to a lifetime dose of 50,000 or in some cases 10,000 milliroentgens. Because so few surface shots had ever been fired, three safety factors were added to Bravo’s: significant fallout was assumed to go on for twelve hours instead of six; particles down to 70 microns were counted instead of 100, which doubled the RADEX; and the sector added on each side for spreading was widened from 10 degrees to 15. These were the margins. They were margins on a source fifteen miles wide.

One more thing had been done months before: the Naval Radiological Defense Laboratory had scaled the fallout measured from Jangle, a 1.1-kiloton surface shot in Nevada in 1951, up to the Castle yields by the cube-root rule. That pattern, like the cylinder, kept the inhabited atolls far from harm, and the task force attached little confidence to it, for the reasons that a desert and a wet reef, a kiloton and ten thousand of them, are not the same thing.

The three briefings

Bravo’s winds were measured from the ships, from the weather islands and from aircraft, and summarised in hodographs, plots of what a balloon would do as it rose through the layers: west in the trade wind up to about 8,000 feet, back east in the “midlevel westerlies” from about 20,000 feet to the tropopause near 60,000, then west again in the stratospheric “Krakatoa easterlies”. Read backwards, the hodograph is the “hotline”, the trace on the surface of where debris from directly over the shot would land as it fell. On the measurement made from the Curtiss forty-five minutes before the shot, the hotline’s closest approach to any inhabited place was 45 miles from Rongelap, for particles starting at 35,000 feet and taking about seven hours to come down.

The briefings were held aboard the Estes at 18:00 on 28 February, at midnight and at 04:30. The five-day advisory had promised a fan from north-west to north-east, and the advisory issued eighteen hours before the shot had told the Pacific command to expect “no significant fallout . . . for the populated Marshalls”. At 18:00 the winds were less favourable; the decision to fire was reaffirmed, with another look at midnight, and a destroyer was moved to a more southerly station. At midnight the winds at 10,000 to 25,000 feet were less favourable still, and those at 20,000 feet “were headed for Rongelap to the east. The predicted speed of these winds was low enough to be of no concern, although it was recognized that both Bikini and Eneman islands would probably be contaminated.” The memorandum of the midnight briefing gives the reasoning as it was made:

“In general, the forecast presented at the midnight briefing was confirmed, except that in the levels between 5 and 15 thousand feet the forecast was light and variable. In an attempt to delineate direction to these winds, it was determined that the best forecast that could be given was for the 10 thousand foot level. This was forecast to be westerly at 10 knots as the most pessimistic situation that could occur. Consequently, the hodograph plot was made using the 10 thousand-foot westerly wind in order to present the most pessimistic situation that would occur. This picture gave resultant winds in the direction of Rongelap and Rongerik, however, it was considered that the distance to Rongelap and Rongerik compared to the resultant wind speeds were such that no fall-out should reach those atolls. From the forecast hodograph the time of travel to Rongelap would have been about 12 to 15 hours. . . . Native atolls in the southeast quadrant were discussed at this point and elsewhere in the briefing. The net result of the forecast was that these atolls should remain favorable due to the forecast long time of travel for fall-out to these places. Specifically, Wotho and Rongelap were considered by name and position, these being the closest native populated atolls in the vicinity of ground zero.”

The outlook recorded was “Bikini: unfavorable; Eniwetok: favorable; Ujelang: favorable, and the native populated atolls in the southeast quadrant from ground zero favorable, since resultant winds in the direction of these areas were considered too slow to move significant fallout to the atolls involved.” Rongelap Island was about forty miles outside the RADEX. Because Bikini itself was expected to be contaminated, the radiological safety office recommended moving the fleet’s sortie area from thirty miles to fifty; the smaller and slower ships went, and the larger ships stayed at thirty to keep voice radio with the firing bunker and a helicopter within reach of the firing party. At 04:30 there were “no significant changes” except that the low winds over Bikini showed more northerly and westerly components, and the shot was confirmed. The 2013 reviewers’ judgement is that the low winds were read correctly: debris at 20,000 feet would have drifted at most twenty miles east before dropping into the trade wind and being carried back to Bikini, which is what the photographs show it doing. The lower part of the cloud “essentially hung above [ground zero]”. It was not the stem that reached Rongelap.

What the cloud did

The fireball was more than three miles across; the cloud rose 45,000 feet in its first minute, through the tropopause, and stopped after five minutes above 115,000 feet. “An immense trumpet-shaped cloud formed. Ten minutes after the explosion, the head of the mushroom cloud had widened into an anvil resting upon the tropopause, measuring 75 mi across and growing rapidly. And from the base of the immense anvil cloud, looking like virga from a desert shower, was debris falling toward the earth below.” Samples from Rongelap held particles up to 500 microns across, seven times the thickness of a hair; the floating collectors near the western edge of the Danger Area caught particles of a millimetre. The large particles had gone up into the stratosphere with the fireball and come straight back down. “The source of the BRAVO fallout was not the expected 15 mi wide cylinder extending upward to the stratosphere, but instead a 100 mi wide cloud resting upon the tropopause.” Clarkson and Graves: “The original source cannot be considered as a point or a relatively small area but must be considered to be an area of about a hundred miles diameter.” Graves, that March: “The theory that a significant fallout does not come from the stratosphere is not substantiated by the facts of BRAVO.”

The conclusion was reached within days. The first detailed analysis, a week after the shot, drew separate fallout envelopes for the stem and for the head of the cloud, taking the head to be at least seventy to a hundred miles across, and the head’s envelope took in the populated atolls. On 2 April the Commission’s general advisory committee was told that heavy particles “falling out of the dome [of the mushroom cloud] . . . effectively enlarged the stem to a diameter of 50-75 miles”. If the preshot RADEX is simply widened by forty or fifty miles, the radius of the stratospheric head, it contains Rongelap and Rongerik.

Two other things the reviewers establish bear on the public story. The first is that the yield, though it “certainly contributed”, was not the cause: the height and shape of a thermonuclear cloud “do not depend strongly on the yield”, the 6.9-megaton Union shot in April rose to 94,000 feet and spread to 45 miles in ten minutes, “not greatly different” from Bravo, and every one of the Castle shots at Bikini put measurable contamination on Rongelap. The second is the winds. On 31 March Strauss said that “the wind failed to follow the predictions but shifted south of that line”, and the Commission’s report of February 1955 repeated the “unexpected shift in the direction of the prevailing winds in the higher altitudes”. The task force had already said otherwise. On 9 April it signalled headquarters that there were only small differences between forecast and actual winds; its weather officers complained on 13 April that they had been “unjustifiably placed in a bad light” and that the chairman’s statement “had been amplified to the discredit of the Air Force Weather Service”; the commander of the Air Force task group wrote in June that “the press releases concerning the 1 March test have misrepresented the quality of the weather forecast.” The Clarkson–Graves memorandum’s finding was that the forecast “was essentially correct”, off by about ten degrees in the significant upper levels, “unfortunately . . . in the wrong direction”, and “well within the normal forecast error.” The 2013 report allows that the winds at 20,000 to 30,000 feet may have shifted enough to put a small part of the stem over Rongelap and Rongerik, and finds that “this effect contributed in no significant way”. Strauss’s shift of the winds, it suggests, came from a briefing of 31 March that described the fallout as a strip twenty miles wide, which would indeed have needed a narrow cloud blown by a shifted wind; the strip was wrong.

There is also the matter of the weather itself. The task force’s charts of winds from 10,000 to 55,000 feet, drawn up afterwards for the whole of February and March, show Bravo and Romeo fired on the two best days of March. They also show 27 and 28 February as better days, when everything was ready; the orders said not before 1 March, and the shot was in fact fired on 28 February by Washington’s calendar. Graves’s letter of 17 April asking that shots be fired when the weather was suitable rather than on a date is the comment on that.

Three maps

How much dust fell where was argued about afterwards in three reconstructions, and the argument is the same argument as the drawing. The Naval Radiological Defense Laboratory, in its Castle report of 1956, spread the radioactivity uniformly through the whole visible cloud: about ninety per cent of the fallout from the 66-mile-wide stratospheric head and ten per cent from the four-mile stem. The RAND Corporation put ten per cent in the stem and most of the rest in the inner lower part of the head, with the particle size varying with height. The Air Force’s Air Research and Development Command, following Lieutenant Colonel Lulejian, whose model had been built on tower shots and airdrops in Nevada, put eighty per cent of the radioactivity in the stem and twenty in the head, with forty per cent of the whole carried by 70-micron particles between 30,000 and 40,000 feet, so that “nearly half of the radiation was thus assumed to come from just 2% of the cloud, and the massive stratospheric cloud head was taken to be essentially free of radioactive debris.”

The three give very different pictures: the Navy’s wide, with the inhabited islands at its edge; the Air Force’s narrow, 170 miles long and 35 wide, with the islands near its axis; RAND’s between. The Air Force pattern is the one that was printed in The Effects of Nuclear Weapons and on the cover of the Congress of Micronesia’s report, and the reviewers say flatly that it is incorrect. The evidence came from the next big shot: for Yankee, 13.5 megatons on 4 May, oceanographers used the sea itself as the collector, sampling the water over a wide area east of Bikini, and the pattern they found matches the Navy’s model and not the Air Force’s. Cloud penetrations in 1956 measured the dose rate in the head of a megaton cloud at five to ten times that in the stem. Glasstone and Dolan wrote in 1977 that “Because of the absence of observations from large areas of the ocean, the choice of the [BRAVO] fallout pattern . . . is largely a matter of guesswork”; the reviewers’ view is that the preponderance of evidence supports the Navy.

Lulejian himself had written after the shot: “I am amazed at the similarity of this event with the fallout at Nevada. As a matter of fact, if the JANGLE Fallout Model had been simply extrapolated to the CASTLE Bravo yield . . . then I’m sure that the CASTLE Bravo shot would not have been fired. This is because all Bikini Atoll islands and the islands of Rongerik, Rongelap, and Bikar are clearly in the fallout area even when such a simple extrapolation is used.” That sentence became part of the lore: that a simple scaling would have shown the danger and the shot should never have been fired. The reviewers’ answer is that the simple extrapolation had been done, months before, and showed nothing of the kind; that Lulejian’s own method, applied with the real winds, puts hazardous fallout from the stem only to the north-west of the inhabited atolls; and that “the error in forecasting the BRAVO fallout was that the contribution from the stratosphere was not adequately considered. Contrary to [Lieutenant Colonel] Lulejian’s widely circulated beliefs on the matter, there was, in fact, significant fallout from the stratosphere.”

Gordon Dunning of the Commission summed it up in 1968: “until 1954 one school of thought held that high yield surface detonations would create intense fallout only in the immediate area of the shot, and that most of the activity would be carried into the stratosphere where it would be scattered widely around the world. March 1, 1954 saw the dismissal of that school — permanently.” The dismissal cost the people of Rongelap their island, and it was the first measurement anyone had made of the thing the drawing had left out.