18

Traffic

The Japanese, now occupying the Solomon Islands to Australia’s north-east, had almost complete dominance of the south-west Pacific beyond Australia. The long-awaited first Allied counter-attack came in August, with landings on two islands in the Solomons: the large central island of Guadalcanal, and the smaller island of Tulagi to its north.

On the morning of 7 August, an American fleet sailed north under the command of Rear Admiral Richmond Kelly Turner, with the First Marine Division under Major General Alexander Archer on board. John Prados described the fleet’s approach in his history of the Solomons Campaign, Islands of Destiny:

The weather closed in as they approached. Low cloud ceilings had kept Japanese search planes away. They might achieve surprise in this, the first major Allied offensive of the war.

With dawn a carefully choreographed sequence began with the order — the first of many times it would be issued — ‘Land the landing force.’ Soon afterward, aircraft from American carriers one hundred miles away swept in to strafe the invasion beaches.1

The Japanese defenders had no idea that the Allies were coming, and were caught completely off-guard. The United States marines landing at Tulagi met no resistance at all, despite having to wade one hundred yards to get to the beach, while the landing force at Guadalcanal met a token resistance before the Japanese defenders retreated to other parts of the island. Tulagi was quickly secured as an Allied base, but the battle for Guadalcanal was only just beginning. It would last for months, with both sides trying to sneak reinforcements to the island without getting their ships sunk by the other side’s planes and submarines.

General Alexander A. Vandergrift and Admiral Turner were faking when they looked at the clouds with crossed fingers. The morning could have brought an unbroken blue sky from horizon to horizon, and yet no reconnaissance planes would have sighted them. They knew this because Pappy Clark, a traffic analyst at Central Bureau, had been charting the Japanese reconnaissance planes over the Solomon Islands, and he knew their daily patterns.2

Since taking part in that first parade at the Caulfield showground two years before as an original member of Jack Ryan’s new special signals unit, Pappy Clark had become an expert traffic analyst. His talent extended beyond applying techniques he learned from others: his creative, music-obsessed mind allowed him to invent new methods and techniques, and to find interesting ways of using the data they had.

Using signals from the air force’s network of direction-finding units in Queensland, Clark plotted the flights of the Kawanishi H8K Flying Boats based much further north at Gasmata, on the island of New Britain. The Kawanishi, nicknamed ‘Emily’ by the Americans, were long-range Japanese aircraft that scoured the Coral Sea for any sign of Allied shipping movements.3

Pappy Clark advised the invasion fleet to wait below the horizon until a certain time, when the Emilys had done their flyover and were heading back to their base. Clark told an ABC interviewer in 1976 that this was the reason he was later given a Member of the British Empire (MBE) award; because, as he put it, the Americans were ‘able to land marines at Tulagi and catch the Japanese at breakfast’.4

The success of the Solomon Island landings did not translate into a quick, bloodless victory. Although the Japanese on Tulagi had been caught by surprise, they mounted a brief but fierce resistance when the marines advanced across the island’s interior.

On Guadalcanal, the easy landings led to jubilation and overconfidence among the green, untested marines, leading them to fall for a simple ruse. A lone, drunk Japanese soldier wandered into the camp, and, when interrogated, told them that the Japanese on the island were starving, weak, and ready to surrender. Vandergrift’s intelligence officer, Lieutenant Colonel Frank B. Goettge, was in charge of the interrogation, and immediately asked Vandergrift to let him go on a mission to find these Japanese soldiers, telling his commander that ‘hundreds of Japanese troops were roaming the jungle west of us ready to surrender’. There was no shortage of marines willing to volunteer. Vandergrift assented, but sent the expedition by boat to a landing point around the coast that was away from known enemy strongholds.

Goettge’s expedition left at night, lost their way, and made landfall at the mouth of a river close to a Japanese encampment. They were surrounded and picked off, one by one, in the darkness. Three of the marines escaped, separately and alone, swimming back to the American beachhead. The last marine to leave the beach was Sergeant Frank L. Few, an American Indian from Buckeye, Arizona. As he swam out in the dark beyond the view of Japanese snipers, his last vision of the beach was the flashes of descending swords as the Japanese soldiers overran the marine’s position.5

In the following days, the main beachhead attracted sustained enemy attack. Meanwhile, the Japanese forces there readied themselves for combat in the overgrown interior. The support fleet came under such fierce aerial bombardment that Admiral Turner moved it away from Guadalcanal’s shores two days later, leaving them stranded without naval support.

This was the beginning of the long campaign for Guadalcanal, a battle that was to drag out for another six months.

Signals were arriving at Central Bureau from Townsville, where 1 Wireless Unit had a team of newly trained Kana operators, as well as a series of direction finders in place across Northern Queensland. The Americans were investigating Japanese army codes, but the volume of messages was still too low for them to make much headway. However, the Australians at Central Bureau, led by Mic Sandford, were putting the output from Townsville to good use in traffic analysis, as Pappy Clark showed with the August landings in the Solomons.

Traffic analysis (known these days as ‘Signals and Network Development’6) was not, as the name might imply, a study of vehicles. It was the investigation of radio traffic: the ebb and flow of the enemy’s radio messages between its military units. Now, as then, a traffic analyst isn’t interested in breaking an encrypted message, or learning what the message says. The analyst is simply interested in the fact that a message has been sent. What can that tell you? With a bit of ingenuity, quite a lot, and it can tell you even more if you look at the patterns across large numbers of messages.

A spike in radio traffic coming from a certain base could mean that large numbers of units were gathering there, or that an operation was about to begin. If an enemy stronghold increased its air-to-ground traffic, that might suggest a build-up of aircraft; the use of tactical codes, by contrast, could suggest coordinated activity with battlefield units.

In 1942, there was a goldmine of information embedded in the preambles to coded Japanese messages.

If a signals officer had a message to send, there was no point in just turning on the transmitter and tapping out the message in Morse or Kana, even if encrypted — the radio broadcast would be picked up by any receiver in range. In the Second World War, the airwaves were full of radio signals from both Allied units and Japanese Axis units, from navies, armies, and air forces, as well as from merchant ships and other units. Nobody would know who you were sending them to, and that included your intended recipient, without sending and receiving identifiers being tagged on. This wasn’t true just for Second World War radio messages — it’s true for all messaging systems. Packages need a mailing address to reach their intended destinations; phone calls can only be made between two unique numbers.

Therefore, every single radio message broadcast by either side contained a preamble — what we today call ‘metadata’. The most important information to include in a preamble was the identity of the sender and recipient. This was done with simple codes known as call signs.

For example, in the naval air-to-ground code, at one point the call sign for the air base Rabaul, to the north of New Guinea, was ‘290’, while the Japanese air base at Dili on Timor was ‘336’. A Japanese signal operator who had to send a radio message from Rabaul to Dili would include the call sign for both bases in the preamble before transmitting the message itself. Thus, the operators at Dili, when their radio receiver picked up the message, would know that it was a message for them, and that it had come from Rabaul.

Depending on what code system was being used, other information in the preamble included the code that the message was encrypted in, the time the message was sent, whether it was a multi-part message, and, in the case of the mainline codes, the unit or individual the message was specifically addressed to.7

Traffic analysis is the forgotten child of signals intelligence, overshadowed by code-breaking, but code-breaking is almost useless without it. For example, there is no point in breaking a code and reading an enemy message that says, ‘We are sending you the 21st Regiment’ if you don’t know who the message was sent from, or who it was sent to. On the other hand, if a traffic analyst tells you that the message was sent from the Japanese commander at Rabaul to the port of Wewak, the message is suddenly quite useful.8

Sinkov, being a mathematician and a leading expert in cryptology, was naturally suited to the code-breaking side of things. He and his expanding cohort of recruits arriving from Washington had cryptanalytical expertise, but no expertise in traffic analysis. Prior to the Second World War, the US army had not engaged in traffic analysis at all. (The US navy had been doing so, however, and did have such expertise, both at Monterey and Hawaii.) The Station 6 officer from the Philippines, Howard Brown, now attached to Central Bureau but currently in Townsville, was the first United States army officer to engage in traffic analysis when he started monitoring air raids against Manila.9 Sinkov’s colleagues at Arlington Hall in Virginia had sent some sigint personnel to England for training in traffic analysis; these personnel had begun some traffic-analysis work at Arlington Hall in April, but it was still a fledgling operation.10

For the Australians, the reverse was true. All of Australia’s cryptanalysis experts — Nave, Room, Trendall, Treweek, Lyons, and a handful of others whom Nave had been training — had been absorbed by the naval operation in Monterey Flats, leaving their compatriots over at Central Bureau with barely any code-breaking expertise. But Mic Sandford’s Australian army contingent at Central Bureau did have expertise in traffic analysis. In the course of travelling across Greece, Crete, Africa, and the Middle East on various campaigns, they had acquired knowledge from the British, and had become proficient in setting up their operation from scratch.

To help them get started with traffic analysis, Newman had supplied Central Bureau with a list of known Japanese army radio frequencies that the Australian navy had collected.

The United States navy also helped. Fabian provided them with a list of known Japanese army call signs, as well as a Japanese army address code that had been broken by his colleagues at Station HYPO in Hawaii.

The American naval sigint analysts were monitoring Japanese naval messages, not army messages. Even so, they had picked up some knowledge of Japanese army call signs, because these call signs would be used in Japanese naval messages during joint operations.

It turned out that, in 1941, the Japanese army had run out of call signs because of its rapid expansion across Asia and the Pacific. It was operating in locations in far-flung parts of the world from Burma, to Singapore, to Timor, and elsewhere — remote places that were not listed in the army’s call sign index. It got around the problem with a quick fix: if there was no call sign for a particular place, operators would simply spell it out using Kana. These were usually phonetic adaptations of the English names of these places. Rabaul, for example, was transmitted as RA BA RU, Sydney was SHI DO NI, and Timor TI MO RO.

This quick fix, with its mixture of proper call signs and on-the-fly spelled-out placenames, had the potential to cause confusion for message recipients. Therefore, if a Japanese army signals officer sent a placename in this way, he would first send two Kana letters, WE RA, and end it with MU WE. Thus, Timor would be sent as the following Kana string: WE RA TI MO RO MU WE.

These were easy to spot in a radio broadcast, because the placenames sent in this way were always bracketed at the start and finish with the Kana letter we. This code was broken in 1941 by Edward Layton, an American naval cryptanalyst. Layton called it the WE-WE code.11

As a token effort to disguise what they were doing, the Japanese army issued a Kana substitution table to be used with these placenames (in which RA became KA, for example). But once Layton realised what they were doing, this was quite easy to unravel.

Although it was only in use for a short time, this address code turned out to be of great value for Sandford’s traffic analysts. The WE-WE code enabled them to identify the location of various units moving into the south-west Pacific. Japanese signals headquarters in Tokyo soon released a new, expanded address-code index that included call signs for all the new locations, so that the WE-WE code was no longer used, but the traffic analysts had used the interval to learn the Imperial Japanese Army’s structure and movements.

Over the second half of 1942, as Japanese and Allied forces clashed in a series of battles in the Solomon Islands and the eastern end of New Guinea, MacArthur’s two main sources of Ultra came from the JN-25, courtesy of the United States navy, and from Sandford’s team of traffic analysts at Central Bureau.

Every transmitted message was of interest to the team. They recorded the preambles, particularly the call signs, before passing the air-to-ground messages to the Australian team and the other army messages to Sinkov.

In June, Booth’s newly operational 1 Wireless Unit in Townsville intercepted a message that used a WE-WE encoded placename, GA DA RU KA NA RU, as the location of an airfield under construction. It did not take long for the traffic analysts to conclude that this probably referred to the island of Guadalcanal in the Solomon Islands. The American forces were already planning an operation in the Solomons, so they expanded it to include a landing at Guadalcanal to stop the airstrip from being finished. Eleven thousand United States marines landed on the island on 7 August. The ensuing battle for Guadalcanal continued for several months until February the following year, when the Japanese forces withdrew.

In July, Sandford reported to Allied commanders that there had been an increase in overall radio traffic from Japanese bases to the North, which he thought strongly indicated that a major operation was imminent, probably within 30 days.12 Based on traffic analysis alone, he was able to advise Willoughby, MacArthur’s chief of intelligence, that the likely target was either Buna, on the New Guinea north coast, or Milne Bay, which was right at the far-eastern tip of New Guinea. Sandford also advised Willoughby that patterns of air-to-ground activity were similar to those observed in the Philippines during the Japanese invasion in early 1942 — an insight almost certainly given to him by Howard Brown from his hectic days at Station 6.

With the recent change in JN-25, many naval messages were now unreadable, and much of Fabian’s advice to MacArthur was also based on traffic analysis. Fabian advised MacArthur that there would be a major landing in New Guinea or the Solomons, and that, based on the movements of troop transports, the initial landings would be consolidated by the development of bases.

The Japanese army duly landed at Buna in July. However, even though the landing had been anticipated, heavy storms prevented the Allied aircraft from harassing them.13

More evidence came in that Milne Bay was the next target. Sandford’s team detected the movement of the Japanese 17th army into the area in readiness for an operation, and a partially decrypted intercept involving them referred to a unit as an ‘occupation force’. MacArthur’s new air force chief, General Kenney, sent reconnaissance planes across to Rabaul and confirmed a build-up. Japanese documents captured by Allied forces also referred to Milne Bay, dovetailing with the traffic analysis.

MacArthur conferred with Blamey about the need to reinforce Milne Bay, where the Australian army had recently established a base. Blamey sent the 18th Australian Infantry Brigade, bringing the total number of Australian soldiers in the bay to about 10,000 troops.

The Japanese army made landfall at Milne Bay on 25 August. Allied commanders waited anxiously for updates from the Australian troops there, but radio messages were infrequent and provided little information. The generals had no idea what was happening.

It was the wet season, and in the hasty, makeshift buildings of the Australian encampment, the army radio sets had got wet, and were out of operation. That was why there were so few update reports.

The Japanese army had landed in the middle of a torrential, lashing storm. Visibility was poor, and neither side was sure what was going on. The battle on the beach and into the sodden coastal jungle was chaotic, with shells ripping through the drenching rain, and soldiers on both sides stumbling through the foliage, hunting each other. The sun broke through and air support arrived, but with soldiers scrambling around in the dense undergrowth below, the pilots could not tell who was on which side.14

In the midst of the communication blackout, MacArthur finally got updates on the progress of the battle, but from naval decrypts supplied by Rudy Fabian. The navy was intercepting Japanese naval messages about the battle, and Fabian was able to inform him that the Australians were winning. One Japanese commander had been killed in the fighting, and another one wounded; only one-third of the Japanese landing force was still operational.

Despite this, the Japanese army established a beachhead, and would not let it go. Sandford discovered reinforcement shipping nearby. Soon afterwards, based on intercepted naval messages, Fabian advised that the shipping in question was on its way to Milne Bay, but that it was in fact an evacuation convoy. Bad weather again prevented planes from attacking the convoy, and the Japanese forces on the bay were evacuated to Rabaul in the north.

The Australian army thwarted the Japanese army’s attempt to capture Milne Bay, in part due to information gleaned by Sandford’s traffic-analysis team. This in no way diminishes the actions of the Australian troops who fought in the Battle of Milne Bay, some of whom were killed in action. The intelligence provided by Ultra was the reason they were there, but Ultra did not fire the guns. The battle still had to be fought on the ground. It was the first time in the Pacific War that a Japanese land-based operation had been successfully repelled.