First Siege of Ulsan

The siege of Ulsan was the first major allied attack on the Japanese castles of the south coast during the second Japanese invasion of Korea (1597–1598), the last phase of the Imjin War. Korean records place the first fighting outside the castle on the 22nd or 23rd day of the 12th lunar month of 1597 and the end of the siege on the 4th day of the 1st month of 1598. Converted with the table of the Korea Astronomy and Space Science Institute (KASI), that is 28 or 29 January to 9 February 1598.

The target was a new castle in the Japanese style (wajō) that the Japanese were building to Katō Kiyomasa’s plan on a hill at Ulsan, at the eastern end of the Japanese line. A Ming army under the civil commissioner Yang Hao and the commander Ma Gui, together with Joseon troops under the commander-in-chief Gwon Yul, attacked before it was finished. They drove the Japanese out of their camps, took the outer earthwork and shut the garrison into the stone-walled castle on the summit.

The castle did not fall. The garrison was short of food and water, and both sides suffered in the winter rain and cold, but assaults on the high stone walls cost the attackers dearly. When Japanese relief forces gathered by land and sea, Yang Hao and Ma Gui lifted the siege; their army was attacked as it withdrew, and Yang later reported the campaign to the emperor as a near success. He was impeached and dismissed in 1598.

In the autumn of 1598 Ma Gui attacked Ulsan a second time, as one of four simultaneous allied offensives. That attack also failed, and Katō left Ulsan only when Japan withdrew from Korea after the death of Toyotomi Hideyoshi. The later sections of this page look at the castle, the siege, the reports about it and what the sources say about numbers.

In the autumn of 1597 the Japanese advance into Korea was halted, and the armies fell back to the south coast. There they kept the fortresses built in 1593 and began a new outer ring of castles; Japanese and Korean reference works list eight, from Ulsan in the east to Suncheon in the west. Ulsan was the easternmost, north of Katō Kiyomasa’s base at Seosaengpo on the coast. The Ming commanders regarded Katō as the strongest of the Japanese generals, and Yang Hao wrote to the emperor that he had to be struck first.

The site was a low, isolated hill at the head of Ulsan Bay, with the Taehwa River estuary along its south side. In 1597 the water reached the foot of the hill, so ships could moor directly below the castle. According to the Encyclopedia of Korean Culture, Katō drew up the plan, Asano Yoshinaga built it with 16,000 men in some forty days from the 12th day of the 10th month, and Ōta Kazuyoshi supervised the work; Japanese accounts date the start of construction to the middle of the 11th month and name Mōri Hidemoto’s troops among the builders, while Yang Hao’s memorial says the castle was begun in the 12th month. The records do not agree on when work started. Stone was taken from the walls of Ulsan town and of the old army headquarters nearby; excavators found a Joseon naval commander’s memorial stele of 1585 reused as a gate stone.

The finished plan had a main enclosure (honmaru) on the summit and two lower enclosures to the north and north-west, with twelve turrets and stone walls built at a slant in the Japanese manner. Around the landward sides ran a long earthen outer wall with a palisade and a ditch, and on the south a landing place for boats. A Ming officer described the castle to the Joseon court as four rings: a low earthen outer wall that could be broken, and three inner rings of stone with buildings along the tops of the walls, from which the defenders could fire down on anyone below.

When the allies attacked, the work was almost but not quite finished. The castle had not yet been stocked with food, and the hill had no well. Japanese accounts say that the Mōri troops had already begun to leave, and that Katō himself was away at Seosaengpo.

Yang Hao and Ma Gui marched south from Hanseong (Seoul) in the 12th lunar month and joined Gwon Yul’s Joseon forces at Gyeongju. To keep Konishi Yukinaga in the west from helping, the Encyclopedia of Korean Culture says, they made a show of moving toward Suncheon. The Ming troops were divided into three divisions; the Joseon commanders included Seong Yun-mun, Yi Si-eon and Jeong Gi-ryong, and the Gyeongsang naval commander Yi Un-ryong was ordered to watch Seosaengpo from the sea.

The allies set out from Gyeongju in three columns in the early hours of the 23rd. Gwon Yul and Yi Deok-hyeong reported that the vanguard of the left division under Li Rumei reached the Japanese camps at dawn, pretended to flee, then turned on the pursuers and took more than 500 heads; other accounts put this first fight on the 22nd. The Japanese camps outside the castle were burned, and the survivors fled up the hill. Katō, at Seosaengpo, came back that night with a small escort; Korean reports say he came on horseback, Japanese accounts by ship.

On the next day the allies attacked the outlying earthworks and the outer wall. Ming troops broke in, burned the buildings and two ships on the river, and the defenders retreated into the stone castle. Ten Ming soldiers got onto the wall and did not come back. The Revised Annals and Jingbirok add that the Ming troops, intent on plunder after taking the outer palisade, did not press on at once, and the defenders shut the gates. From then on every assault failed: the walls stood on a steep hill, the defenders fired through loopholes, and the Ming cannon, as a Joseon official later told the king, overshot the walls because the castle stood so high.

The allies then surrounded the hill, hoping that hunger and thirst would finish the garrison. They also tried to burn their way in: troops were sent forward carrying bundles of brushwood and shields to pile at the foot of the walls. The Joseon commanders Yi Si-eon and Seong Yun-mun reported a second large attack on the 27th (2 February), and wrote that almost all the Korean soldiers from Gyeongju and Ulsan who had carried brushwood for the fire attack were killed, and that the people of six or seven districts, including Andong, had been exhausted by carrying grain for the army.

The castle had been surrounded before it was supplied. A Ming officer told the Joseon court that there was little food inside and the water had run out. Yu Seong-ryong, who visited the allied camp on the 29th, wrote in Jingbirok that Dosan had no water, so the Japanese slipped out every night; on Yang’s orders Kim Eung-seo set an ambush by a spring outside the castle and over several nights caught more than a hundred men, all emaciated by hunger and barely alive.

Prisoners told Yang Hao that the garrison numbered fewer than 3,000, that the bodies of men killed by cannon and arrows or worn out by hunger and thirst lay everywhere, and that the 200 musketeers were each receiving one hop (a small measure) of raw rice a day. The Encyclopedia of Korean Culture adds that the defenders drank the blood and urine of their horses and boiled paper and the earth of the walls for food; Japanese Wikipedia repeats this, citing the encyclopedia. These stories come from later compilations and testimony and cannot all be checked; several accounts describe severe hunger and thirst, although its scale and the particular details remain uncertain.

One story about Katō himself reached the Joseon court in the 3rd month of 1598. A Korean sailor who had been a prisoner at Seosaengpo said that Katō had entered the castle with fifty men, twenty of whom were killed on the way, and that when things became desperate he twice tried to kill himself and was stopped by his officers. This is hearsay and is not found in the Japanese accounts used here; Japanese Wikipedia, citing Japanese sources, says Katō entered with about 500 men.

The besiegers suffered too. The Revised Annals say that heavy rain was followed by bitter cold, so that the soldiers were exhausted and frostbitten and many horses froze to death. Jingbirok describes the troops’ hands and feet cracked with cold in the winter rain, and Yu Seong-ryong reported that the men were camped in the open as an east wind brought rain. The allies had expected to finish the siege quickly and were not ready for a long one: a Joseon official later said that in the first rush the Ming troops had dropped much of their equipment and had no siege engines when they needed them.

While the castle held, the other Japanese commanders gathered at Seosaengpo. According to Japanese accounts, Yamaguchi Munenaga (for Kobayakawa Hideaki), Kuroda Nagamasa, Mōri Hidemoto, Hachisuka Iemasa, Nabeshima Naoshige and his son and Katō Yoshiaki arrived one after another, and a fleet including Chōsokabe Motochika came by sea. Joseon reports already noted more than forty Japanese ships anchored on the lower Taehwa River, and Yang Hao posted 2,000 Zhejiang infantry and 1,000 cavalry on the bank to stop them landing.

There were also talks. Yang Hao’s memorial and Korean reports say that Katō, through interpreters, offered to surrender, or to withdraw if he were allowed to leave as Konishi had once been, and that Yang refused; Japanese accounts say that the allies sent envoys to demand surrender, and that Katō used the talks to gain time. The sources agree only that letters passed and that no agreement was reached.

On the 2nd day of the 1st month (7 February) Japanese troops appeared with many banners on distant hills facing the castle, and on the 3rd they came down to skirmish. That night the allies made a last assault by torchlight from all sides, from midnight until dawn. According to Yi Si-eon and Seong Yun-mun, the Japanese fire fell like rain, losses were very heavy, and no one reached the wall.

On the 4th (9 February) Yang and Ma Gui decided to withdraw. The infantry was sent off first while the cavalry covered the rear, and stores and gunpowder left behind were burned. The Zhejiang troops guarding the ford at Jeontan were not told and were caught by the Japanese coming down from the hills; few of the infantry escaped, and many threw away their armour to get away. Li Rumei’s Joseon liaison officer reported that the Japanese pursued for thirty li, that Ming dead were “some say 3,000, some say 4,000”, and that the retreating troops plundered villages, raped village women and even killed people on their way north. Yang withdrew to Hanseong, leaving Ma Gui and Gwon Yul to hold Gyeongju.

The first news to reach Hanseong was triumphant. Messengers spoke of hundreds of heads taken and of Katō trapped in his “earthen caves”, and King Seonjo thanked the Ming commanders for the victory. But Yi Deok-hyeong, the Joseon official attached to Yang Hao, reported that Yang would not allow the heads to be counted until the castle had fallen and told him not to report the allies’ own losses.

After the retreat Yang sent the emperor a memorial that told the story as a near success. His troops had stormed the Japanese positions for forty li, he wrote, killing uncounted numbers by fire and drowning and beheading or capturing more than 1,300; Katō had been on the point of capture when a relief army of tens of thousands arrived, and he had withdrawn to save his tired men. The imperial reply praised his courage and the many heads taken, but asked why he now offered to resign instead of planning to finish the war. Joseon’s own reports were less flattering: Yi Deok-hyeong counted nearly 700 Ming dead and over 3,000 wounded during the siege, not including the men lost at Jeontan in the retreat.

In 1598 the Ming war ministry official Ding Yingtai, who served on the staff in Korea, impeached Yang. The Ming History, compiled under the Qing, says that the army’s rolls showed almost 20,000 dead, that Yang suppressed them and reported only a hundred or so, and that Ding listed 28 crimes; King Seonjo spoke of 29 capital charges. Other charges, as reported in Seoul, were that Yang had made peace with Katō and that he had encouraged Joseon to build fortresses. Yang was dismissed and replaced by Wan Shide; news of the dismissal had reached Seoul by the 23rd day of the 6th lunar month (25 July 1598).

The Joseon court defended him. Its officials blamed quarrels between the Ming northern and southern troops over credit for the battle, and the king sent the envoys Choe Cheon-geon and then Yi Won-ik to Beijing to plead for Yang. Jingbirok says the king saw him off in tears when he left. Ding then accused the Joseon king himself of conspiring with Japan, a charge the Joseon court now had to answer. Yang Hao later returned to office and commanded the Ming army defeated by Nurhaci’s Jurchens at Sarhu in 1619; he was executed in 1629.

In the autumn of 1598 the Ming supreme commander Xing Jie tried again with a larger plan: four simultaneous attacks. Ma Gui was to take Ulsan in the east, Dong Yiyuan Sacheon in the centre and Liu Ting Suncheon in the west, while the admiral Chen Lin, with Yi Sun-sin’s Joseon fleet, attacked from the sea. Toyotomi Hideyoshi had in fact died on 18 September 1598, but his death was being kept secret.

According to the Encyclopedia of Korean Culture, Ma Gui led about 24,000 Ming troops and Kim Eung-seo 5,500 Koreans out of Gyeongju on the 21st day of the 9th month (20 October), and the attack on the castle began on the 22nd (21 October); a Korean report summarised in Japanese Wikipedia has the fighting at the castle begin on the 20th. Kim struck first at the Japanese near Dongnae to cut the road between Ulsan and Busan; Ming troops cleared the Japanese post at the Onjeong hot springs and brought out more than a thousand Korean captives. A Ming messenger reported that Ma Gui had burned large Japanese grain stores across the river and that over 1,100 Koreans, probably including these captives, had come over to his lines.

This time Katō was ready. The same messenger said that the Japanese had dug a moat and let sea water into it so that no one could cross, and that the castle looked impossible to take. Japanese accounts say that Katō kept his men inside and refused to be drawn out. On the night of the 22nd a Japanese raid killed five Ming soldiers. Ma Gui’s liaison officer soon reported that after falling back from the inner castle the commander had lost heart and wanted to retire to Gyeongju.

The news of Dong Yiyuan’s disastrous defeat at Sacheon on the 1st day of the 10th month (30 October) settled it. On the 4th (2 November) Ma Gui’s infantry withdrew with all its baggage and weapons, leaving only cavalry at Gyeongju. The Encyclopedia of Korean Culture dates the fighting at Ulsan to the 22nd–25th of the 9th month, while Korean reports show the army staying near the castle until early in the 10th month. Katō burned the castle and left on the 18th day of the 11th month (15 December 1598), as the Japanese withdrew from Korea; Ma Gui’s troops then entered it.

On 17 January 1598, less than two weeks before the siege began, Tsar Feodor I of Russia died without an heir. With him the ruling line descended from Ivan the Terrible, the Rurikid tsars, came to an end; his brother-in-law Boris Godunov became tsar, and Russia slid toward the crisis known as the Time of Troubles.

In France, King Henry IV signed the Edict of Nantes in April 1598, granting the Protestant Huguenots substantial rights and helping to end the French Wars of Religion. In May France and Spain made peace at Vervins, and in September King Philip II of Spain died at the age of 71.

None of these events had anything to do with Ulsan. Placing them side by side shows only that 1598, the last year of the war in Korea, was also a year of endings and settlements elsewhere.

After the war the burned castle was not abandoned at once. According to Korean Wikipedia, the Joseon navy used the site for a time, and from 1624 it held a shipyard for building warships, which the Encyclopedia of Korean Culture says was moved to Gaeunpo in 1654. In the late Joseon period the hill came to be called Hakseong (“crane castle”), a name that was later attached to the castle itself.

In 1913 a local notable, Kim Hong-jo, bought the hill and began planting it with black pines, cherry and plum trees to make a park, which later passed to the local authorities of Ulsan. It is now Hakseong Park, one of the city’s best-known parks. The paths, flower beds and other facilities built for the park, Korean Wikipedia notes, destroyed much of the north-western walls and the lower enclosures.

The castle’s heritage designation has changed several times. Designated a historic site under Japanese rule, it was made Historic Site No. 9 of the Republic of Korea in 1963 under the name Ulsan Hakseong, removed from the national list in 1997 together with other Japanese castles, and on 30 October 1997 designated Ulsan Metropolitan City Cultural Heritage Material No. 7 under the name Ulsan Waeseong (Ulsan Japanese castle). Excavations in 1986, 2011 and 2015 have traced the enclosures, a section of wall on the plain and the east gate.

Today the stone walls of the main enclosure on the summit can still be seen, while the river bank where ships once moored lies under houses and the estuary has moved far to the east. In Kumamoto, the city of Katō Kiyomasa’s later castle, a tram stop called Ulsan-machi is said to take its name from Ulsan or from people brought from there.

The allied army is usually put at 40,000–60,000 men. The Ming History speaks of 40,000 Ming troops in three divisions; a Japanese general-staff history, followed by Japanese Wikipedia, lists about 56,900 men including 12,500 Koreans; Korean Wikipedia gives 40,000 Ming and 11,500 Koreans; English Wikipedia speaks of 50,000. The Encyclopedia of Korean Culture gives 44,000 Ming and over 50,000 under Gwon Yul, a figure much larger than the others. At the start of the siege Yang Hao and Ma Gui said that only 3,500 Korean soldiers had actually arrived.

Estimates of the garrison vary even more. A Ming officer told the Joseon court that some 10,000 Japanese were inside, and the Japanese account Kiyomasa Kōraijin oboegaki, as cited in Japanese Wikipedia, gives 10,000; Yi Deok-hyeong and Gwon Yul thought there seemed to be about 2,000, and prisoners told Yang fewer than 3,000. Later Japanese works give 20,000 or 23,000, which Japanese Wikipedia considers too high. The numbers present may also have changed as stragglers and workers were driven into the castle.

Allied losses during the siege are best documented. Yi Deok-hyeong’s count, in a report dated the 6th day of the 1st month, gave nearly 700 Ming dead and more than 3,000 wounded, with the losses in the retreat still uncounted; a liaison officer reported rumours of 3,000 or 4,000 Ming dead in the retreat. English Wikipedia gives 1,621 dead and 2,908 wounded according to Ming sources. Against these stand much larger figures: the Ming History’s “almost 20,000”, and the 10,386 bodies that a document of the Asano family says were counted around the castle. Korean losses, which the Joseon commanders described as very heavy among the troops of Gyeongju and Ulsan, are not totalled in the sources used here.

Japanese losses are uncertain. Allied reports claimed more than 500 heads on the first day, 800 on the next, more than 1,300 beheaded or captured according to Yang Hao and about 690 for Li Rumei’s division; the Encyclopedia of Korean Culture gives 661 heads. These were claims made in a contest for credit and cannot be checked. The number of Japanese who died of hunger, thirst and cold inside the castle is not known.

Most of what we know comes from Joseon reports in the Veritable Records of King Seonjo: dispatches from Gwon Yul, from the officials attached to the Ming commanders (Yi Deok-hyeong with Yang Hao, Jang Un-ik with Ma Gui, Yi Deok-yeol with Li Rumei) and from provincial commanders, as well as Ming messengers questioned at court. They were written within days or weeks, but each writer saw only part of the field, and the date heading an entry is the date of recording or of the court’s discussion, not necessarily the date of the events described. Yang Hao’s own memorial to the emperor is preserved in the same records. Yu Seong-ryong, who visited the siege, gives a brief account in Jingbirok.

The Ming History, compiled under the Qing dynasty, gives a strongly critical account of Yang Hao, and it should be compared with the contemporary Joseon reports. Its story that Yang called off a successful attack by Chen Yin so that his friend Li Rumei could take the credit, and its figure of 20,000 dead, are not found in those reports. Joseon officials, who admired Yang, blamed the quarrels between Ming factions and gave a different account of who had fought best; Yi Deok-hyeong told the king that the troops blamed Yang for sounding the retreat after the first day’s victory.

Japanese sources include clan records and later war tales, summarised in Japanese reference works. Some are very unreliable: one later chronicle claims an allied army of 800,000 and puts Kobayakawa Hideaki at Ulsan, although he was not there. The monk Keinen, a physician-monk in the service of Ōta Kazuyoshi, one of the commanders in the castle, kept a diary of the 1597–1598 campaign (Chōsen nichinichiki). Japanese Wikipedia mentions it in passing, but this page has not checked its Ulsan passages and does not rely on it.

Even the dates need care. Joseon, Ming and Japan all used lunar calendars, but not always the same one. Korean records place the first fight on the 22nd or the 23rd of the 12th month; Japanese accounts give the 22nd, which Japanese Wikipedia converts as 29 January 1598, the same day as the Korean 23rd. This page uses the Joseon dates converted with the KASI table. English Wikipedia’s end date of 19 February does not match the 4th day of the 1st month given in the Korean records, which is 9 February.

Ulsan shows how hard the Japanese coastal castles were to take by storm. Built on steep hills, with layered enclosures, slanting stone walls and buildings from which musketeers could fire down, they suited defence by musketeers, whom the Japanese had in large numbers. Joseon observers reported that the allied cannon could not hit the castle on its height, that men carrying brushwood were shot down before they reached the walls, and that the soldiers who first got onto the wall did not return. Even unfinished, the castle held.

It also shows the weakness of such castles. Ulsan had no well and had not been provisioned, and within days the garrison was starving and desperate for water. The Korean Wikipedia article on the castle says that Katō remembered this when he later built Kumamoto Castle with many wells; the sources used here do not document that connection directly. In the autumn of 1598, with the castle complete, a sea-water moat dug and the garrison prepared, the allies could make no headway against it.

Above all, Ulsan depended on the sea. Ships could moor below the castle; Japanese boats from Seosaengpo came by water every day during the siege, and Ming reports say they were turned back; but Japanese ships still gathered in the river, and the relief army came by land and sea together. The allies could surround the hill on land, but they could not stop the Japanese from bringing ships up to it; the Joseon navy had been almost destroyed at Chilcheollyang in 1597. In the end the castles of Ulsan, Sacheon and Suncheon were all still in Japanese hands when Japan decided to withdraw, and the last great battle of the war was fought at sea, at Noryang.

The siege had consequences in Japan as well. Japanese accounts say that thirteen commanders, alarmed by Ulsan, proposed to Hideyoshi that Ulsan, Suncheon and Yangsan be abandoned; he refused. His inspectors reported that Hachisuka Iemasa and Kuroda Nagamasa had failed to fight in a preliminary clash during the relief, and both were reprimanded and confined to their domains; the punishments were reversed in 1599 after Hideyoshi’s death. One interpretation, cited in Japanese Wikipedia, links the resentment this caused among commanders such as Katō against Ishida Mitsunari, a relative of one of the inspectors, to the split that led to the Battle of Sekigahara in 1600.