A Japanese Imperial Navy signal intercepted on February 18, 1943, carried three letters that would decide the fate of thousands of men: RZM. A clerk typed the designator onto a routing slip and filed the decrypt without ceremony. Six days later, the convoy that signal described sat on the bottom of the Bismarck Sea, and the Imperial Japanese Army never again sent a troop convoy through that strait. The decrypt arrived at the headquarters of General George Kenney, commander of Allied Air Forces in the Southwest Pacific Area, on February 23.
A note attached said the original had been retranslated for 17 hours to confirm a date inside the text. The convoy counted from that retranslation: eight transports, eight destroyers, and the better part of an infantry division embarked. Kenney had been waiting for exactly that formation since the previous autumn, when a smaller one had slipped through and put troops ashore at Lae. The arithmetic that sent the formation through was sound on paper.
Eight transports under destroyer escort could land 3,000 troops at Lae in two days of steaming. The same passage in January had drawn 416 Allied sorties for the loss of two transports, and the troops aboard those ships had been picked up by escorts and landed at Lae anyway. The Imperial General Staff in Tokyo priced the risk against the reward and called it an even bet. That arithmetic missed a measured American hit rate.
In the six weeks before the convoy sailed, the Fifth Air Force had used a wreck aground in Port Moresby Harbor as a target. Pilots in the light and medium bomber groups dropped 30 to 40 bombs into the exercise at mast height in pairs until a well-trained crew was expected to strike the target nearly every time. The January figure was the old rate. The March convoy would meet the new one.
Rabaul to Lae is 600 nautical miles of open water drawn through the Bismarck Sea and the Vitiaz Strait, with a wide opening and a narrow exit. A convoy making seven knots takes three days to clear it. There is one approach in and one approach out, and both can be watched from a single coastwatcher station on the south coast of New Britain. The first day went to the Japanese plan.
On the afternoon of March 2, high-level strikes found the formation south of Cape Gloucester and put one transport under. The escorting destroyers drew together, recovered the survivors, and pressed on at the same speed. By the standard Tokyo’s staff officers had priced, the formation was now under budget: one transport of eight gone, troops redistributed to the other seven, the formation intact, the destination still two days away. Japanese convoy officers, in interrogations conducted by American intelligence at Yokohama in September 1945, later described what happened on March 3 as unlike anything they had seen.
The first thing they could not fit to prior experience was the altitude at which the bombs were released. The Japanese anti-aircraft battery of 1943 was built around a height-finding system that worked by optical rangefinder and a mechanical computer, and a height-finding system needs a height. A formation at 15,000 feet gives the rangefinder a long, slow target to measure. A formation coming in at 50 feet gives it nothing.
The bombs that struck the convoy that morning were released at mast height, between 30 and 50 feet above the water, by aircraft flying at 250 to 275 miles per hour. The rangefinder could not measure that altitude, and the computer could not compute against it. The guns were left to point and fire by eye. The second novelty was that the bombs arrived sideways.
Mast height release is not the same as skip bombing, though the two are often run together in secondary literature. In skip bombing, an aircraft drops a bomb with a delayed fuse at low altitude and a shallow angle; the bomb bounces across the water like a flat stone and strikes the ship’s side below the armor belt. In mast height bombing, the bomb is released at the same altitude with no intention of skipping, on a flat trajectory that carries it directly into the side of the hull. Both techniques were used on March 3.
Japanese convoy officers described the bombs as arriving on a line and striking the side of the ship rather than falling onto the deck. They had no precedent for it. A wreck had been aground in Port Moresby Harbor since 1923, and the Fifth Air Force had been using it as a target since the previous November. The third surprise was that the bomber shot first.
The standard medium bomber of 1941 mounted one or two forward guns of rifle caliber in the nose and did its real work with the bomb load inside. A factory-fresh Mitchell carried a single half-inch gun in the nose and a second in the top turret. The nose gun was normally fired by the bombardier in self-defense. The Mitchells that came over the formation on March 3 carried 10 forward half-inch guns in the nose, all fixed, all fired by the pilot through a single trigger.
The paired drill was that the front aircraft of two would open fire at a thousand yards from the target and put its bullets onto the bridge and anti-aircraft positions. The wingman held his fire and put his bombs onto the hole. The guns’ role was to suppress the ship’s anti-aircraft fire for the six seconds between the opening of fire and the release of the bomb. The deputy commander of one Japanese transport, asked at Yokohama how many guns he estimated a single American bomber carried, answered eight.
He said eight had been half enough. The fourth problem was mechanical. The standard Japanese anti-aircraft gun on a destroyer of 1943 was the 25-mm Type 96, mounted in pairs or triplets, trained by hand, with a 15-round magazine that had to be changed by a loader standing on the platform. The published rate of train for the standard mounting sat between four and six degrees per second.
A Mitchell crossing the convoy at 250 miles per hour on a beam attack was moving across the field of fire at about seven degrees per second at the moment of release. A gun that cannot be brought onto the target inside the time available is furniture. The Japanese Navy had built its short-range anti-aircraft defense on a gun whose train rate and magazine capacity were both fixed in the late 1920s against aircraft moving at 150 miles per hour. The gun had not been updated since.
The pause while the magazine was changed was the moment a gun captain on the Taimae Maru could see the next pair of bombers already in their run. The fifth surprise was the coordination of altitudes. Allied aircraft came in at three layers. Above, at 8,000 to 10,000 feet, heavy bombers dropped onto the deck and superstructure.
In the middle, at 3,000 to 5,000 feet, medium bombers flew conventional formation, forcing gun crews to track upward. At mast height, at 50 feet, pairs of strafing Mitchells put guns and bombs onto the hull. A Japanese anti-aircraft gunner had to choose which of three altitudes to track. Whichever he chose, the other two were on him.
What happened on the water between the second run and the eighth is not easy to set down. The destroyers Arashio, Asashio, Tokitsukaze, and Shirayuki were hit and stopped in succession. The transports went in their own sequence. The reasons the Allies gave at the time for strafing runs against survivors in the water were operational and stated in plain terms inside the after-action summaries.
A soldier rescued from the water and put back into the line at Lae was a soldier still in the war. A soldier who was not rescued was not. The Allies attacked the destroyers recovering survivors, and the destroyers stopped recovering them. The sixth failure was the convoy’s fighter cover, which never appeared.
One hundred Imperial Japanese Army and Navy fighters had been allocated to the convoy’s air defense, drawn from fields around Rabaul and Wewak, and they were to be over the convoy at first light on March 3. The allocation was made on the Army-Navy coordination circuit two days before sailing. The reason it failed was a single design fault that ran through the whole Japanese command structure. The Army and the Navy operated different radio fits on different frequencies with no common net.
The message ordering the cover to be airborne at dawn went out on the Navy circuit and arrived on the Army’s circuit too late. Convoy officers stated in interrogations that they had expected cover, had looked for it, and had seen it arrive in small numbers and at the wrong altitudes. Whether Japanese gun crews turned their guns onto those friendly fighters, mistaking them for American aircraft, is an Allied inference, not a Japanese one. The seventh problem was that the enemy already knew the formation’s approach.
The RZM designator had been broken in the decrypts, and the course had been pulled from the signal. But the course had also been beaten by a weather forecast. The 18th Army headquarters had ordered the convoy to take the northern approach through the Bismarck Sea on the strength of a meteorological assessment calling for low cloud cover and poor visibility across the strait on March 3. The cloud cover did not arrive.
The forecast rested on data from a station at Wewak that had not been updated in four days, and the staff officer who signed it had signed the same assessment for the previous two days without checking the source data. The cloud came in six hours after the formation had been attacked. The whole defense had rested on a weather forecast that was 12 hours out of date when it was issued, and the convoy officers had not been told. Most accounts of this battle stop at the tactics because the tactics are the visible part.
What decided it was a few specifications, a mounting drawing, and a signal intercept. Those sit in ordnance reports and decrypt files rather than in battle narratives. The ordnance reports go backward in time from the convoy. In December 1942, at the depot at Townsville, Major Paul Gunn supervised the rebuilding of a single Mitchell bomber into a gunship carrying a 75-mm cannon in the nose.
Gunn, an American officer who had been an airline pilot in peacetime, was known to his crews as Pappy. The bombardier’s position was cut away and replaced by the cannon and a manual loader. The aircraft was intended to fly alongside a formation of standard Mitchells, fire one round into the side of a ship from a thousand yards, and break off to allow the formation to deliver its bombs. The door behind the nose wheel fell off every time the cannon fired, because the recoil mount was welded to a frame not built to take a recoil load.
The prototype flew in this condition for six weeks. The cannon variant saw no service over the Bismarck Sea. What it left behind was the field engineering practice that had produced it, and the next application of that practice was the 10-gun nose. In October 1942, the Eglin Field Proving Ground in Florida issued a report on the minimum altitude attack of waterborne surface vessels.
The arithmetic in the report: a bomb released at 50 feet and 250 miles per hour on a flat trajectory would strike the side of a ship one and a half seconds after release. The standard American instantaneous fuse carried a delay of one-tenth of a second between impact and detonation, meaning the bomb would go off inside the hull above the waterline. The Eglin report had set the lower fuse delay for mast-height attack at four seconds, calculating that a bomb entering the water at 50 feet and 250 miles per hour would continue forward for four seconds before slowing to a stop. A fuse set at four seconds would detonate the bomb at a depth of about 12 feet below the armor belt.
The Australian fuses available to the Fifth Air Force in October 1942 carried a lower delay of five seconds. The four-second figure was not achievable with the Australian fuse. The depot at Townsville cut the delay by removing half of the delay element and recrimping the fuses in batches of 50 in the weeks before the convoy sailed. Earlier modifications had laid the groundwork.
In September 1942, Gunn supervised the rebuilding of a single Havoc light bomber at the depot in Brisbane with four half-inch guns welded into the nose where the bombardier’s position had been. The modification was not written up in any Army Air Force paperwork. The aircraft was assigned to the 89th Bombardment Squadron and flown on operations within 10 days. The guns came from wrecked Airacobra fighters written off at the repair depot, and the mounts were fabricated from mild steel cut to size at the depot.
No engineering drawings survived. The aircraft’s first operational sortie was a strafing attack on a Japanese barge off Buna on September 11, 1942. The barge was sunk. In July 1942, at Nadi on the island of Viti Levu in Fiji, Kenney’s aide Major William Benn borrowed a single Mitchell from the 92nd Bombardment Squadron and flew it out over coral knobs at 50 feet with five practice bombs of 100 pounds each loaded in the bay.
The exercise was unofficial. Three of the five bombs skipped on the surface. The other two entered the water on a flat trajectory and came to rest on the bottom within 15 feet of the coral. The group commander of the 92nd wrote a memorandum to Kenney saying the heavy bomber was not intended for low-level work and that the exercise should not be repeated.
The exercise was repeated with a different aircraft two weeks later. The origin of the principle sits in a classroom at Langley Field, Virginia, in the late 1920s. The Air Corps Tactical School taught in its 1928–1929 curriculum that a bomb released from a low level on a flat trajectory would strike a vertical target at the same angular displacement as a shell from a naval gun. The mathematics of the trajectory was the mathematics of ballistics, not of bombing.
The principle had been on the books for 15 years before it was used. What changed at the Bismarck Sea was not the principle. What changed was that a commander who had a use for it had finally found a convoy worth using it on. The final arithmetic runs like this.
The Fifth Air Force and the Royal Australian Air Force put 210 sorties over the formation on March 3 against an allocation of 250. American bomb expenditure was 278, of which 83 went as direct hits. The convoy had carried 7,000 troops. About 1,200 reached Lae.
The gun crews spent 20,000 rounds of 25-mm anti-aircraft ammunition against a published average for the Type 96 of 1,500 rounds per aircraft shot down. The gun was credited with two American aircraft destroyed in the action. The hit rate for mast height release on March 3 was assessed at 35 percent. The hit rate for medium-level attack on the same morning was 7.
5 percent. The hit rate for medium-level attack on the same formation the previous October was 3 percent. The communiqué issued by the Imperial General Staff in Tokyo on March 5 stated that the convoy had sustained minor losses and that the troops had been landed. The communiqué was withdrawn on March 7.
The count of troops lost in transit ran to 20,000 taken across the period of the campaign from January to March 1943, a count made at the headquarters of the 18th Army at Rabaul, not at the formation’s destination. An entire division was put on ships by men who had done the arithmetic, priced the risk honestly, and been right about everything except the last 12 weeks of somebody else’s engineering.