The Castle That Became the Tokyo Imperial Palace

On the third day of the third lunar month of 1868 — 3 April by the Gregorian calendar, not the 11 April sometimes cited by sources that confuse the meeting with the formal handover a month later — Katsu Kaishū sat down with Saigō Takamori in a small house in Takanawa, south of the city, and surrendered Edo Castle. Between them on the mat they had an agreement that the Tokugawa shogunate would evacuate the castle without a fight, Tokugawa Yoshinobu would go into house arrest at Kan’ei-ji, and the imperial army — 50,000 men who had been preparing to burn the city — would walk in through the gates instead of smashing them. The conversation saved a million people. It also saved the walls, the moats, the stone foundations, the three surviving yagura, and the site the Meiji government would move the Emperor onto eighteen months later.

That is the load-bearing fact for everything else in this article. Edo Castle was never stormed. It was never shelled. It never surrendered at swordpoint.

The shogun simply walked out, and the imperial household walked in, and the buildings that were lost after 1868 were lost to fire and to bureaucratic indifference — not to war. What you can walk through today is the same ground Tokugawa Ieyasu laid out in the 1590s, with most of the same outer stonework, a handful of the same yagura, and the great empty tenshu-dai platform that has stood bare since 1657. The Emperor of Japan lives on a castle footprint — the only reigning monarch in the world who does.

Aerial view of Tokyo Imperial Palace and surrounding grounds on the former Edo Castle site in central Chiyoda ward
The footprint from above. What you are looking at is the inner bailey of Edo Castle — Honmaru at top-centre, Ninomaru to its east, Nishinomaru south where the modern palace sits, and the Fukiage gardens sweeping north-west. The moat system you can still see is the second ring; the outer ring, which ran from Iidabashi to Yotsuya to Akasaka to Tameike, was filled in between the 1870s and 1910 for the railway and the barracks. The whole Imperial compound today is 1.15 square kilometres. Ieyasu’s castle was roughly 20. Photo: Aerial photograph by the Ministry of Land, Infrastructure, Transport and Tourism, Attribution.

What Edo Castle actually is

Edo Castle — Edo-jō (江戸城), also called Chiyoda-jō (千代田城) — is the largest castle complex Japan ever built. At the sōgamae perimeter it enclosed 2,082 hectares, which is about five times the area of Osaka Castle and 1.7 times the modern Chiyoda ward. Inside that perimeter ran three concentric moat rings connected in a clockwise spiral, twenty baileys, thirty-eight gates, twenty yagura turrets and a five-tier tenshu that until 1657 was the tallest wooden tower ever built in Japan at 58 metres over a 10-metre stone base.

The inner bailey — Honmaru, Ninomaru, Sannomaru, Nishinomaru, Fukiage, Kitanomaru — is the walled compound you can still walk through today. The middle moat is mostly intact. The outer moat survives in fragments: you can see it at Ichigaya, at Yotsuya, and at Akasaka-mitsuke where the Benkei-bori ponds still hold water. Everything else within the original 15.7-kilometre outer circuit has been built over. Tokyo’s financial district sits on Ieyasu’s lower-bailey daimyō mansions; Marunouchi is literally “inside the round” of the outer moat.

Scale model of Edo Castle inner bailey showing honmaru tenshu palace layout moats and yagura towers in the Kan'ei period
Scale model of the inner bailey in its Kan’ei-era configuration (c.1640), showing the five-tier keep at the centre-left, the Honmaru palace complex filling most of the flat ground, and the small Ninomaru quarters to the east. The model is to scale but the height of the keep as shown is, if anything, understated — the real thing at 58 metres of superstructure plus 10 metres of stone base was closer to the height of the current National Diet Building. Photo: Mc681, CC BY-SA 4.0.

The keep is gone. The inner and middle moat walls are the originals. The tenshu-dai — the stone platform the keep stood on — is still there, empty, 45 metres on a side, rebuilt by Maeda Tsunanori of Kaga in 1658 after the fire. You can climb it. You can stand on top of it and look south across the Honmaru lawn and understand how the shogunal world was laid out. The Kaga-rebuild platform is six-ken (10.9 m) shorter than the Kan’ei original because they never expected to put a tower back on it.

Stone platform or tenshu-dai of Edo Castle main keep in Imperial East Garden with no tower on top
The tenshu-dai. The Kan’ei keep burned on 2 March 1657 and no keep has stood here in the 369 years since. The white granite was cut in Kaga — in what is now Ishikawa Prefecture — and floated down by Maeda Tsunanori’s crews the summer after the fire. The platform is accessible and free; a set of steps on the north face takes you up. Go up it. Nobody does, because they are looking for a building, and there is no building. Photo: AMANO Jun-ichi, CC BY 3.0.

Ōta Dōkan and the founding of Chiyoda-jō, 1457

The first castle on this ridge was not Ieyasu’s and not Hōjō Ujitsuna’s — it was Ōta Dōkan’s, built in 1457 for the Ōgigayatsu Uesugi during the Kyōtoku Disturbance. Dōkan was a monk-warrior of the kind the Muromachi era kept producing: literate, a poet, a siege engineer, fatally useful to his lord until he was too useful and the lord had him killed. His Edo Castle was a three-bailey hirayama-jō — “flat-hilltop castle” — built on the eastern lip of the Kōjimachi plateau, where the Hibiya inlet came in close to the base of the ridge and the seabound travel and the inland travel met.

Portrait of Ōta Dōkan the Muromachi era warrior poet monk who founded Edo Castle in 1457 for the Ōgigayatsu Uesugi
Ōta Dōkan in the portrait held at Daiji-ji, painted by the 18th-century Ōta descendant Ōta Sukenobu. Dōkan is not the most famous Muromachi figure, but he is arguably the single most consequential city-founder in Japanese history — Edo-Tokyo is the capital it is because he chose this ridge. He was assassinated on his lord’s orders in 1486 at Itsukaichi in Sagami, which is roughly what happens to useful men in every century.

The castle Dōkan built was a three-kuruwa structure — Shijō, Chūjō and Gejō in the period’s Chinese-influenced naming — connected by kiri-gishi cut-away slopes and water moats. He lived in the main bailey in a residence called Seishō-ken, with a high pavilion behind it for viewing the bay. There was a plum orchard on the north side where the 10th-century exile-scholar Sugawara Michizane had been enshrined, which is the remote reason there is still a Yushima Tenjin a few kilometres north today. None of the physical structure remains. But the spiral layout Dōkan used is the spiral Ieyasu’s engineers extended 140 years later.

Dōkan survives in Tokyo culture through one story: the yamabuki incident. Caught in a squall on a hunt in what is now Shinjuku, he knocked on a farmhouse door to ask for a straw raincoat and was handed instead a branch of yellow yamabuki rose by a girl who said nothing. Dōkan left angry.

Later that night someone explained to him the Heian-era poem she had been quoting — “Seven-fold, eight-fold the yamabuki blooms, but it is sad that they have not a single fruit” — where the pun on mi means both fruit and rain-cape. Dōkan, who had prided himself on his literature, had missed it; he spent the rest of his life writing waka. The story is almost certainly apocryphal, and it is also the only Muromachi anecdote every Tokyo schoolchild knows.

Ukiyo-e woodblock print of Ōta Dōkan receiving a yamabuki rose branch from a village girl in the Yamabuki no sato legend
Yūsai Toshiaki’s late-Meiji rendering of the yamabuki incident. Dōkan is the figure on horseback at right, drenched and confused; the girl is offering the branch. The modern Yamabuki-no-Sato marker is in Bunkyō ward, about four kilometres north of the Imperial East Garden — it is a stone under a tree, and you would not find it without directions. The picture has had more staying power than the site.

After Dōkan’s death in 1486 the castle passed to the Ōgigayatsu Uesugi main line, then in 1524 to the Odawara Hōjō under Hōjō Ujitsuna, who made it one of the six branch castles that held the eastern Kantō. The Hōjō held Edo for 66 years, through the rise of Takeda Shingen and Uesugi Kenshin further north, through Oda Nobunaga’s unification push, and into the 1590 Odawara Campaign that ended them. They kept Edo in good order but did not substantially extend it. When Toyotomi Hideyoshi took Odawara and distributed the Kantō to Tokugawa Ieyasu in August 1590, Edo Castle passed to Ieyasu as a down-at-heel secondary fort with a collapsing stockade and a hundred-odd thatched houses around it. That is the castle Ieyasu walked into.

Ieyasu arrives, 1590

Ieyasu entered Edo on the first day of the eighth lunar month of 1590 — 30 August by the Gregorian calendar — and the sight that met him was, by every account, unimpressive. The Dōkan-era residence was 133 years old and rotting. The moats were silted. The castle town was a fishing village with about a hundred straw-thatched cottages clustered near the north gate. The Hibiya inlet came right up to the eastern wall, which meant that at high tide the lower baileys were literally in the sea. Ieyasu had just traded Suruga, Tōtōmi, Mikawa, Kai and Shinano — five provinces, his ancestral homeland — for this.

Portrait of Tokugawa Ieyasu the first Tokugawa shogun who made Edo Castle his Kantō seat in 1590 by Kanō Tan'yū
Tokugawa Ieyasu in the formal Kanō Tan’yū portrait held at Osaka Castle. Ieyasu was 47 when he took Edo — old by Sengoku standards to start what amounted to a new province. He spent the following ten years not fortifying the castle (that came after Sekigahara) but building the town: draining the Hibiya inlet, cutting the Dōsan-bori canal, bringing Sumida water down through Kanda, and running the first Nihonbashi bridge as zero kilometre of all five great highways.

Hideyoshi’s politics were simple. Ieyasu was too powerful to destroy but dangerous left where he was; the Kantō transfer was a gilt exile disguised as promotion. What Hideyoshi did not count on, and what Ieyasu understood immediately, was that the Kantō plain was a better base than Mikawa ever could be.

It was bigger. It had more arable land, a harbour that could feed a capital, and was one ridge-crossing from the Tōhoku that Hideyoshi had just acquired but not yet integrated. Ieyasu spent the first ten years drainage-engineering the harbour and the plain. The castle could wait.

What he did expand, in those pre-1600 years, was the bailey layout. The old Dōkan three-bailey plan got extended with a new Nishinomaru (west), Sannomaru (third), Fukiage (the upper gardens) and Kita-no-maru (north). The spoil from the Hibiya drainage went to fill in the land between the castle and the bay, which is why everything east of the Imperial Palace down to modern Shinbashi sits on 1590s reclamation. The Dōsan-bori canal, named for the Ieyasu physician Nakurai Dōsan who supervised its cutting, still runs under Nihonbashi as a concrete box culvert. The water is there. The original banks are gone.

Keichō era map of Edo Castle showing the baileys and moat system in the early 17th century after Ieyasu's initial expansion
A Keichō-era (1596–1615) picture-map of Edo Castle — almost certainly copied later from an earlier original — showing the inner bailey, the Nishinomaru west enclosure, and the gates as they stood at the transition into full Tenka-bushin construction. The circular pond at bottom-left is the Hibiya inlet, still partly open sea. Within ten years everything east of the main keep on this map would be reclaimed land.

Tenka-bushin: the shogunate builds its castle, 1603-1636

After the Battle of Sekigahara in October 1600, Ieyasu had the political authority to stop building the castle with Tokugawa money and start building it with everyone else’s. The device he used was tenka-bushin (天下普請) — “realm-wide public works” — a forced-contribution system where the ~200 daimyō of Japan were assigned quotas of stone, timber, men and money to deliver to the Edo site on pain of losing their domains. The first big round ran from 1603 to 1607 and it produced the castle’s basic form. The second ran from 1618 to 1622 under Tokugawa Hidetada. The third, 1628 to 1636, under Tokugawa Iemitsu, closed the outer moat ring and finished the sōgamae.

The assignment lists survive and they make sobering reading. Hosokawa Tadaoki, Maeda Toshinaga, Ikeda Terumasa, Katō Kiyomasa, Fukushima Masanori, Asano Yukinaga, Kuroda Nagamasa, Tanaka Yoshimasa, Nabeshima Katsushige and sixteen others drew the outer stone walls. Kuroda Nagamasa alone took the tenshu-dai. The stone itself mostly came by sea from Izu — Ieyasu’s engineers preferred Izu granite for its density — with the largest blocks reportedly up to 24 tonnes, cut at the Atami and Itō quarries and rafted 100 kilometres up the coast on purpose-built stone barges.

Ōte-mon gate of Edo Castle and Imperial East Garden reconstruction of the main eastern entrance
The Ōte-mon as it stands today is a 1968 reconstruction; the outer masugata gate-box and the lower stone course are original. The corner bastion is the work of Katō Kiyomasa’s Higo masons during the 1606 round of tenka-bushin — the stones are identifiable by the “+” tally mark Kiyomasa used on his quarried blocks. His Kiyomasa-no-Ido well, still flowing in the Honmaru, is the more famous of his Edo relics. Photo: RSSFSO, CC BY-SA 4.0.

Kiyomasa’s Edo contribution is the one that people still point to. He was the most highly-regarded castle mason of his generation — Kumamoto is his work, and the Nagoya stone bastions — and for the Edo job he sent a team of Higo stone-cutters who produced the Ōte-mon bastion, the Fujimi-yagura platform, and the Honmaru well that still bears his name. The well, Kiyomasa-no-Ido, is in the Honmaru lawn at the base of the tenshu-dai. It still holds water. The story that Kiyomasa dug it himself is at best decorative, but the well-cap stones carry his masons’ tally marks, so it is at least a Higo-team job. You can see it on the East Garden loop today.

The forced-contribution system is the same administrative logic that would later produce sankin-kōtai, the alternate attendance system that kept the daimyō in permanent rotation between Edo and their home domains. Tenka-bushin was the prototype: a tax levied as labour and material, payable to the shogunate, non-negotiable. The Tokugawa ran it for 33 years on Edo Castle alone. By the time Iemitsu finished the Sanada-bori outer moat in 1636 the daimyō had collectively shifted, at a conservative estimate, 1.2 million tonnes of stone from Izu to the Kantō — enough to resurface most of the Tōkaidō twice over.

Detail from Edo-zu Byōbu screen painting showing panoramic view of Edo Castle and city in the early 17th century
A detail from the Edo-zu Byōbu, the pair of six-panel folding screens now held at the National Museum of Japanese History in Sakura, showing Edo and the castle roughly as they stood in the Kan’ei era around 1634. The five-tier white keep is the Kan’ei-do tenshu that would burn in 1657. The surrounding clusters of white-walled daimyō mansions are the kami-yashiki of the inner bailey — and they are essentially a to-scale rendering, not an idealisation.

Three keeps in thirty years, and then no keep for ever

The Kan’ei-do keep that appears on the byōbu was the third main tenshu to stand on the site, not the first. Ieyasu had Kuroda Nagamasa throw up the Keichō-do (慶長度) in 1607 — a white-walled five-tier tower, lead-tiled, roughly 48 metres tall, with the central pillars of Kiso fir running full-height. Hidetada pulled it down in 1622 and rebuilt it as the Genna-do (元和度), shifted to a more northerly position in the bailey, six-tier inside with an enshrinement of the Tōshō-gū to Ieyasu on the interior. Iemitsu pulled that one down in 1636 and built the third, the Kan’ei-do (寛永度) — black copper tiles, copper sheet on the walls, 58 metres of superstructure, the tallest in Japanese history. Thirty years of keeps, three generations, no continuity of design.

Architectural elevation drawing of the Kan'ei-do tenshu of Edo Castle from the 1630s plans
Reconstructed elevation of the Kan’ei-do keep based on the Kōra family carpenters’ plan set held at the Kunaichō archives. Note the copper-tiled roof and the absence of ornamental karahafu on the lower tiers — this was a late-style layered-type keep, much plainer than the elaborate Azuchi-Momoyama towers at Himeji and Matsumoto. At 58 metres of superstructure it would have dominated the skyline in a way no building in Edo did again until the 1920s. Illustration: DasMurmeltier, CC BY-SA 3.0.

The Kan’ei-do lasted twenty-one years. On the 18th day of the first lunar month of 1657 — 2 March by the Gregorian calendar — a fire starting in a Honmyō-ji temple in Hongō, three kilometres north of the castle, caught a winter gale off the Musashino plain and ran south. By the evening of the third day the fire had killed an estimated 100,000 people, burned 60 per cent of Edo including most of the inner bailey, and taken the Kan’ei-do keep down with it. The contemporary accounts say a copper shutter on the second tier was left open by a careless guard, a flying ember caught the interior, and the building burned from the inside out because the copper exterior was sealing the flames in. The two great pillars crashed through all six floors.

Woodblock illustration of the 1657 Meireki fire destroying Edo city and the castle keep from the Musashiabumi chronicle
The Meireki fire as illustrated in Asai Ryōi’s Musashiabumi chronicle, published a few years after the event. The building on fire at centre-right is the Kan’ei-do keep of Edo Castle; the figures running south in the foreground are trying to reach the Sumida river. Ryōi’s death toll of 107,000 is the basis for most modern estimates. No single disaster in Japanese history before the 1923 earthquake killed more people in a single city.

Maeda Tsunanori of Kaga rebuilt the tenshu-dai during the summer of 1658, using Kaga white granite, to a height of six ken (10.9 metres) instead of the original ten. Plans for a fourth keep were drawn up and materials quietly ordered. Then — and this is the turning point of the whole Edo Castle story — the shogunal elder Hoshina Masayuki intervened.

Masayuki argued that Nobunaga had introduced the tenshu at Gifu (which he had) and Azuchi, that it was essentially a symbol of warlord self-promotion, and that a castle’s defences did not require one. The whole project was paused to prioritise city reconstruction. It was never unpaused — every subsequent revival attempt (Arai Hakuseki in the 1710s, the Meiji government in the 1870s, two NPO campaigns in the 21st century) has died on the same argument.

The decision is, on reflection, extraordinary. The Tokugawa had three generations of tenshu on file. The plans from the Kan’ei rebuild survive today — the “Edo-jō tenshu tate-wari zu” at the Kōra family archive is a complete set of full-elevation and floor plans, detailed enough that modern woodworkers could rebuild the keep from it. They chose not to. For 366 years they have chosen not to. The shogunate found it politically more useful to be a regime without a keep — an administrative bureaucracy whose symbolic centre was deliberately empty — than to replace the tower. The Fujimi-yagura, on the south ridge, became the de facto keep for ceremonial purposes and remains so today.

Fujimi-yagura three-tier corner turret of Edo Castle on the original 1659 stone base seen from the Imperial Palace grounds
The Fujimi-yagura, on the south-east corner of the Honmaru, is the three-tier turret that served as the de facto keep from 1659 onwards. The superstructure is a 1659 rebuild after the Meireki fire took the original; the stone base is Keichō-era and the oldest continuous structural element still visible on the castle site. The name means “Mount Fuji-viewing turret”, and on clear winter mornings from the top floor — closed to the public — it does. Photo: Reggaeman, CC BY-SA 3.0.

Edo peace, 1657-1867

Between the Meireki fire and the Meiji Restoration the castle’s political function stayed roughly constant. Fifteen Tokugawa shoguns governed Japan from here. About 5,000 men were on the premises during working hours — bugyō in the administrative offices, bangata guards on the rotating three-shift garrison, koshō pages in the residential quarters — and another 600 to 1,000 women lived permanently in the Ōoku, the inner chambers no man under the rank of shogun or physician could enter. The total full-time population of the castle on any given weekday was around 6,000. It was the largest single workplace in Japan for 264 years.

The daimyō came in and out on the sankin-kōtai rota, presenting themselves in the Ōhiroma on designated days. Senior rōjū elders served one month on, three months off, commuting from mansions in what is now Marunouchi. Metsuke — the shogunate’s inspectors — reported from their offices in the Sannomaru. On ceremonial days the Ōhiroma held 500 daimyō at once, arranged by rank in a sequence of tatami rows that was worked out to the centimetre. The social geometry of the Edo political order was encoded in the Honmaru floor plan.

Reconstructed scale model interior of Ōhiroma grand hall of the Honmaru palace of Edo Castle with daimyō in formation
Scale model of the Ōhiroma (Grand Hall) in the Honmaru palace, the ceremonial chamber where daimyō assembled for audience with the shogun. The space was tiered — the shogun sat raised on the jōdan platform at top, senior tozama daimyō on the chūdan below, and the rest in descending order of rank. The ukiyo-e of “Matsu-no-Ōrōka” — the pine-painted corridor outside — is the setting of the 1701 Asano/Kira incident that started the 47 Rōnin. Photo: Daderot, CC BY 3.0 DEED.

Security on the outer ring was the responsibility of specific fudai retainers’ houses. The Sakurada-mon gate was routinely assigned to the Ii clan of Hikone — the house of Ii Naomasa, the original Sekigahara-era Red Devil. The arrangement held for 260 years. It ended on the snowy morning of 24 March 1860 when Ii Naosuke, the shogun’s chief councillor and direct descendant of Naomasa, was cut down outside Sakurada-mon by seventeen Mito and Satsuma rōnin. The assassination broke the shogunate’s grip on foreign policy and started the sequence that led, eight years later, to Katsu and Saigō in Takanawa.

The Edo-period peace was also the peak of Edo’s fires. Twelve catastrophic conflagrations touched the castle between 1657 and 1867 — the Ninomaru in 1747, the Honmaru and Ōoku in 1844, the Nishinomaru in 1852, another Honmaru fire in 1859, and a combined Honmaru-Ninomaru-Nishinomaru loss in 1863 that ended the reconstruction cycle entirely. The Honmaru palace was never rebuilt after 1863. The Ninomaru was never rebuilt after 1867. By the time the Meiji imperial household arrived in November 1868 the physical plant of Edo Castle was substantially smaller than the plant Iemitsu had left in 1651.

The bloodless surrender, 1868

The Boshin War started in the Kyoto suburbs on 27 January 1868 at the Battle of Toba-Fushimi and it went, for the Tokugawa side, very badly very quickly. Tokugawa Yoshinobu, the fifteenth and last shogun, retreated to Edo aboard the warship Kaiyō-maru arriving on 12 February. By mid-March the imperial Eastern Expeditionary Force under Arisugawa Taruhito was moving east along the Tōkaidō toward Edo with 50,000 men, Armstrong guns and Minié rifles; the plan was to take the castle by force and reduce what was left of Tokugawa power at a stroke.

Katsu Kaishū was the Tokugawa navy minister — 45 years old, San Francisco-returned (he had captained the 1860 Kanrin-maru mission), a reformer who had spent the previous decade arguing that the shogunate should modernise or fall. He was the one member of Yoshinobu’s inner circle who both accepted the impossibility of military resistance and had standing with the imperial side. The imperial commander on the Edo front was Saigō Takamori of Satsuma, 41 years old, who had been running the war since Toba-Fushimi. Katsu had known Saigō through Sakamoto Ryōma for several years. That personal relationship is the hinge of everything.

Memorial stone marking the site of the Katsu Saigō meeting where Edo Castle surrender was negotiated in 1868
The marker at the site of the first Katsu-Yamaoka-Saigō meeting in what is now Takanawa, Minato ward. The building is gone; what stands here is a 1941 monument in a small compound next to a Keihin-Tōhoku line rail bridge. It is easy to miss — the signage is in Japanese only, the compound is shared with a car park, and tourist maps do not feature it. Tamachi or Shinagawa Station, ten minutes’ walk. Worth the detour for the quietness more than for the monument itself.

Katsu and Saigō met in Takanawa on 13 March (old calendar; 5 April Gregorian) and again on 14 March (6 April Gregorian). The earlier emissary — Yamaoka Tesshū, Katsu’s deputy — had already travelled through the imperial lines to Sunpu in mid-March to deliver the preliminary terms to Saigō personally. At the Takanawa meetings the two men worked out the full surrender: Yoshinobu would retire to Mito, the castle would be evacuated, imperial forces would enter without resistance, and the Tokugawa fleet and armoury would be handed over. The formal transfer of the castle took place a month later, on 3 May 1868 by the Gregorian calendar, when Tayasu Yoshiyori presented the keys in the Nishinomaru to the imperial representatives.

Painting by Yūki Somei depicting the surrender of Edo Castle to imperial forces in 1868 held at Meiji Memorial Picture Gallery
Yūki Somei’s 1935 painting of the surrender, held at the Meiji Memorial Picture Gallery in Gaien. Saigō is the bearded figure on the left; Katsu is the smaller man in Tokugawa robes on the right. The setting is partly imagined — the historical Katsu-Saigō meetings were in a domestic interior at Takanawa, not the open hall Somei painted — but the picture is the one that fixed the scene in the 20th-century Japanese public imagination. Every schoolbook reproduction is ultimately a copy of this.

Sources routinely confuse the three dates. The usually-cited “11 April 1868” is the Gregorian conversion of the meeting on 14 March (old calendar), which was an agreement in principle. The formal surrender was 3 May 1868 Gregorian. The lunar-calendar 3rd day of the 3rd month (the canonical “Edo Castle surrender date” in Japanese patriotic historiography) corresponds to 25 March by modern reckoning and refers to the first Katsu-Saigō contact via Yamaoka. Take your pick; the practical outcome is the same. A million people did not die, the castle stones stayed in place, and the 260-year Edo peace got a final non-violent coda.

From Edo-jō to Kōkyo, 1868-1945

The Meiji emperor entered the castle on 26 November 1868, renamed it Tōkei-jō (東京城), and — having briefly returned to Kyoto — came back to stay on 9 May 1869 when it was renamed Kōjō (皇城, “Imperial Castle”). The Nishinomaru palace, formerly the shogunal residence, became the imperial residence. The Honmaru, wrecked by the 1863 fire, stayed as an open lawn. The Ninomaru, also burned in 1863, was left as gardens. The changing nomenclature across the period — Tōkei-jō, Kōjō, Kyūjō (宮城), and from 1948 Kōkyo (皇居) — is not a rebrand but a step-by-step secularisation of a warrior fortress into an imperial household and then into a public garden.

Felice Beato 1860s photograph of Edo Castle Nishinomaru palace known as the Tycoon Palace before the 1873 fire
Felice Beato’s 1860s photograph of the Nishinomaru palace — what the foreign residents of Edo called the “Tycoon’s Palace” (from taikun, the term the shogunate used in its treaty correspondence). This is the last full photographic record of the main shogunal residence before the 1873 fire took it. Beato’s Yokohama studio produced most of the surviving 1860s views of Edo Castle; his single best series is at MIT’s OpenCourseWare archive, free to view.

On the night of 5 May 1873 a fire destroyed the Nishinomaru palace, taking with it most of the shogunal residential structure Beato had photographed a decade earlier. The Meiji household lived in the Akasaka detached palace for fifteen years while the new Kyūjō palace was built on the Nishinomaru footprint. The new palace opened in 1888, and from 1889 to 1945 it was the working residence of the Meiji, Taishō and early Shōwa emperors. It was a wooden building, designed in a conscious fusion of Japanese exterior and European interior — coffered Japanese ceilings over European parquet floors, traditional roof lines with Belgian chandeliers underneath. It held 737 square metres of main audience hall, the largest single Japanese-style interior of the 19th century.

Felice Beato photograph showing wooden Nijūbashi double bridge over the Edo Castle moat in the 1860s before stone replacement
Beato’s view of the original wooden Nijūbashi (“double bridge”) over the Nishinomaru moat, before the Meiji government replaced both spans in stone and iron in 1887. The name nijūbashi originally referred to the two-storey wooden bridge on the far span — the structural innovation that gave it the name — and was retained for the iconic paired-arch stone-and-iron replacements that stand there today.

The Meiji palace burned on the night of 25 May 1945 during the B-29 firebombing raid on central Tokyo. 88 of the 101 wooden structures on the imperial compound were destroyed. Pilot Richard Lineberger’s squadron later recorded that the palace had been a secondary target on a 29 July raid where they dropped 2,000-pound bombs specifically into the grounds; the full-scale destruction was the 25 May fire. The imperial family had moved into the Obunko underground annexe earlier that year, which is where Emperor Hirohito held the 14 August privy council meeting that concluded the Pacific War. The surrender was, effectively, decided in a bomb shelter under the old Fukiage garden.

From 1948 the compound was renamed Kōkyo (皇居, “Imperial Residence”), losing the last of its (城, “castle”) nomenclature. The architect Junzō Yoshimura designed the current Kyūden palace, completed in 1968 on the Nishinomaru footprint — steel-reinforced concrete with traditional gabled-hip roof lines and a large-span audience hall. The Fukiage residence, where Emperor Naruhito lives today, was built by Shōzō Uchii and finished in 1993. The Honmaru, Ninomaru and Sannomaru were opened to the public in 1968 as the Imperial East Garden.

What actually survives, and what does not

There is no keep. There has been no keep since 1657, and the argument that the Tokugawa and then the Meiji and Shōwa governments made against rebuilding it — that an administrative centre does not require a symbol of warlord power — has held for 369 years. You should not come to Edo Castle expecting a Himeji-style wooden tower or a Matsumoto-style black keep. There is no Kumamoto-scale donjon here either. There is a lot of moat, a lot of stone wall, three original yagura, one original gate, and the empty tenshu-dai.

The three original yagura — Fujimi-yagura, Fushimi-yagura and Tatsumi-yagura — are the oldest wooden superstructures on the site, all rebuilt after the Meireki fire in the 1660s. The Fujimi is on the Honmaru south ridge, the de facto keep substitute. The Fushimi sits above the Nijūbashi moat on the southern approach — this is the yagura you see in every tourist photograph of the palace. The Tatsumi is in the Ninomaru south-east corner, less visible. All three are Important Cultural Properties. None is open to the public interior.

Fushimi-yagura corner turret of Edo Castle above the inner moat visible from the Nijūbashi viewing area
The Fushimi-yagura, seen from the Nijūbashi view-point on the southern approach. The name comes from the tradition that the timber was salvaged from Fushimi Castle in Kyoto after it was demolished in 1623 — although the actual structural core is 1660s Meireki-rebuild work with possibly some Fushimi material in the upper storey. It is the yagura that appears on the back of the 1,000-yen Bank of Japan note series D (1950). Photo: Kamoseiro, CC BY-SA 3.0.

The original gates that survive are the Sakurada-mon and, in modified form, the Tayasu-mon and the Shimizu-mon. The Ōte-mon is a 1968 reconstruction above original stone. The Hirakawa-mon and the Kita-Hanebashi-mon are substantially original Edo-period structures. The Sakurada-mon is the gate where Ii Naosuke was cut down in 1860 — the outer masugata courtyard, where the assassination happened, still stands, and the small modern marker inside notes the date without much commentary.

Sakurada-mon gate outer masugata courtyard of Edo Castle where Ii Naosuke was assassinated in 1860
The Sakurada-mon outer gate. The masugata courtyard you pass through — the quadrilateral box between the outer gate and the inner one — is the space in which Ii Naosuke was dragged from his palanquin and beheaded on the snowy morning of 24 March 1860. The gate is a kōrai-mon / watariyagura pair and is now a National Important Cultural Property. The stonework on the outer bastion is original 1606 Keichō tenka-bushin, with the “ten” tally marks of Katō Kiyomasa’s Higo masons visible on the foundation course.

The stone walls are more important than the buildings. The inner-moat wall, the Honmaru revetments, the tenshu-dai, the Kiyomasa-no-Ido, the Momiji-yama retaining wall, the Sakurada outer bastion and most of the eastern inner wall are all original Keichō-to-Kan’ei-era masonry, 1606-1636. If you know the tally-mark system — each daimyō’s teams carved an identifying mark on their quarried stones to prove delivery for the quota — you can walk the Honmaru perimeter and identify which section was Hosokawa, which was Kuroda, which was Ikeda Terumasa, which was Kiyomasa. The marks are small, about the size of a 500-yen coin, and worn. But they are there.

Where to visit Edo Castle today

Imperial East Garden (Kōkyo Higashi-Gyoen)

The East Garden occupies the old Honmaru, Ninomaru and Sannomaru compounds and is the only part of the inner bailey fully open to the public. Entry is free. The main gate is the Ōte-mon on the east side — Ōtemachi Station exit C13a, about 200 metres.

The garden is open Tuesdays, Thursdays, Saturdays and Sundays year-round, and also Mondays and Wednesdays if they are public holidays; closed December 28 through January 3. You get a numbered token on entry and hand it back on exit — this is how they count visitors. No reservation required; allow two to three hours for the full circuit.

Imperial East Garden of Tokyo showing Honmaru lawn and historic walls open to public on former Edo Castle inner bailey
The Honmaru lawn, looking north from just below the Fujimi-yagura toward the tenshu-dai on the far side. The garden-design sensibility here is restrained — deliberately so. You are walking on the site of the Honmaru palace, which was the largest wooden building in Edo-period Japan (35,000 square metres of floor area); every visible gesture of the current landscape is about what is not there. Japanese garden design in the karesansui tradition works through absences; this one does it at scale. Photo: DXR, CC BY-SA 4.0.

Things to see, in order: the Ōte-mon itself with its 1606 stonework and 1968 superstructure; the Hyakunin-bansho guard-house on the right as you enter; the Ōban-sho beyond it; the Nakanomon stone base; the Kiyomasa-no-Ido well at the foot of the tenshu-dai; the tenshu-dai itself, climbable; and the Honmaru lawn running south to the Fujimi-yagura. Loop back through the Ninomaru garden — a rare surviving Edo-period kaiyū-shiki stroll garden, recreated in 1968 from Edo plans — and exit via the Hirakawa-mon on the north side. The full walk is about 2.5 kilometres. Bring water in summer; there are no vending machines inside the main compound.

Nijūbashi and the south front

The Nijūbashi — or properly, the Nijūbashi approach — is the set of twin bridges on the southern face of the palace, where every tourist photograph of the Imperial Palace is taken. The front span is the Seimon Ishibashi (“main gate stone bridge”), a two-arch granite structure built in 1887. Behind it, barely visible from the public viewpoint, is the Seimon Tetsubashi (“main gate iron bridge”), the 1964 replacement for the original Meiji-era iron span. The Fushimi-yagura stands directly above the inner bridge.

Nijūbashi two arch stone bridge and Fushimi-yagura turret on the south front of the Tokyo Imperial Palace
Nijūbashi from the Kokyo-mae Hiroba plaza, the standard public viewpoint. The stone arches in front are the 1887 Seimon Ishibashi; the turret behind is the Fushimi-yagura; the partly-visible iron span behind the stone bridge is the modern 1964 Seimon Tetsubashi. Get here an hour before sunrise if you want the reflection shot without crowds — between 10am and 3pm the plaza is solid with tour groups. Photo: Kirin Square, CC BY-SA 3.0.

The Seimon Ishibashi is not normally open to the public. Twice a year — New Year’s general audience on 2 January and the Emperor’s Birthday on 23 February — the inner Nakamon gate is opened and the public is allowed across both bridges into the Kyūden Totei plaza, where the imperial family appears on a balcony. The Emperor gives a short speech of about 90 seconds. I have been to the New Year audience twice. The crowd is in the tens of thousands and the atmosphere is closer to a political rally than anything sentimental. Worth doing once if you happen to be in Tokyo on 2 January; not worth planning a trip around.

Sakurada-mon and the Ii Naosuke site

Sakurada-mon is the western gate on the southern moat, closest to Sakuradamon Station (exit 3) on the Yūrakuchō line. It is a full surviving masugata gate-pair from the 1620s Hidetada-era rebuild, including both the outer kōrai-mon and the inner watariyagura, and the courtyard between them is where Ii Naosuke was killed on 24 March 1860. There is a small modern bronze plaque inside the gate marking the date — nothing more. The Ii family‘s gate-guard assignment ran from 1603 to the very morning of the assassination; Ii Naosuke was on his way in to the castle, from his Hikone mansion on the other side of the bridge, when it happened.

The gate is open at all hours and free. You can walk right through it and stand in the courtyard; no signage, no guides, no tourist concession. The main Hikone clan mansion across the moat is gone — it was on the land now occupied by the Ministry of Justice building — but the Ii’s second residence survives as the Kyūchū Sanjō-den ceremony hall on the current palace grounds. The combination of the gate, the plaque, the 1860 date and the geometry of how the assassins positioned themselves (in front of and behind the palanquin simultaneously, on the corner where the road turned into the courtyard) makes the whole site one of the most politically dense few hundred square metres in Tokyo.

Chidorigafuchi and the north moat

The Chidorigafuchi moat on the north-west side of the castle is the most photographed stretch of Edo-period water in Tokyo — about 700 metres of boat-rentable moat, lined with cherry trees, running from the Tayasu-mon at the top to the Chidorigafuchi green-way at the bottom. The trees are somei-yoshino, planted in the early Shōwa period on a site that was moat-bank from 1636. In peak bloom in late March or the first days of April the rowing boats are stacked four-deep at the rental pier and the water is not visible for petals.

Chidorigafuchi moat cherry blossom and rowing boats at Tokyo Imperial Palace in the peak hanami week
Chidorigafuchi in peak hanami week. The trees are 1930s plantings on a moat-bank that has been here since 1636. If you come for the blossom, rent a boat on a weekday morning — the rental pier opens at 9.30 and by 10.30 the queue is an hour long. The sakura reference here is the somei-yoshino that dominate every Tokyo hanami site since the cultivar was standardised in the 1880s. Photo: Mstyslav Chernov, CC BY-SA 4.0.

Combine the Chidorigafuchi loop with a walk through Kitanomaru Park, which occupies the castle’s old northern enceinte and contains the Nippon Budōkan (the wooden octagon built for the 1964 Olympics), the National Museum of Modern Art, and the Science Museum. The Tayasu-mon at the top of the park is an original Edo-period gate, 1685 rebuild after a storm, one of the three survivor gates on the north side. It is the one the Tayasu Tokugawa family — one of the three gosankyō branches — used as their compound entrance; Tayasu Yoshiyori, who handed over the castle keys in 1868, came out through it on his way to Kan’ei-ji.

The tenshu-dai, which nobody climbs

The tenshu-dai — the empty keep platform in the Honmaru — is the feature most visitors miss because they are looking for a tower. Go up it anyway. The stone staircase on the north face, installed in the 1968 East Garden redesign, takes you to the top of the 10.9-metre platform; you stand on a 45 × 41-metre granite rectangle with nothing on it. From the south-east corner you can see the Fujimi-yagura; from the north-east corner, the Hirakawa-mon. The platform is the single tallest surviving stonework on the castle site.

The experience of the empty platform is the thing to bring away from Edo Castle. The space is the point. Japan’s other great early-modern castles — Himeji’s renritsu-shiki cluster, Matsumoto’s black keep over its pine moat, Inuyama‘s perched-on-the-Kisogawa tower, Kumamoto’s Kiyomasa-built double keep — kept their tenshu or rebuilt them. Edo’s decision, twice over, to not rebuild was a deliberate political statement in the 1650s and a deliberate repetition of that statement in every subsequent generation. You can argue the Tokugawa were pacifist in their way. The absent keep is the argument.

What to read, and what to skip

The canonical English-language book on the castle is William Coaldrake’s Architecture and Authority in Japan (Routledge, 1996), which has a long and accurate chapter on Edo Castle’s symbolic architecture. For the surrender specifically, Romulus Hillsborough’s Samurai Tales and his longer Ryoma: Life of a Renaissance Samurai treat the Katsu-Saigō negotiations in more depth than most summaries. For the Meireki fire, Asai Ryōi’s 17th-century Musashiabumi chronicle is translated; the modern study by Seidensticker in Low City, High City (1983) uses it well.

Skip the guidebook treatment of the “Imperial Palace” that does not mention the castle history. You are not visiting an imperial palace; you are visiting a Tokugawa shogunal fortress that the imperial household moved into in 1868 without substantial reconstruction. The asymmetry between what the signage emphasises (Meiji palace, imperial garden, current emperor) and what is actually physically there (Ieyasu’s moats, Hidetada’s stone walls, Iemitsu’s gate bastions, a 1658 Kaga granite platform where a 1636 keep used to stand) is the interesting thing. The signage is going to mostly not help you.

One last thing

If you are at the Takanawa monument on a weekday when the commuter trains are sparse, or at the tenshu-dai on an off-season afternoon, the site becomes quiet enough to hear the logic of what happened in 1868. Katsu walked in with no army and a signed letter from a shogun who was finished. Saigō walked in with 50,000 men, Armstrong guns, and instructions to take the city. They agreed in a small room that nothing would burn.

The whole weight of that decision is present in the castle as it stands today — in the walls that never got shelled, in the empty keep platform that was not rebuilt to commemorate a victory, in the turrets that survived because they had nothing to survive. Go to the Takanawa marker first if you can, and then walk north along the old Tōkaidō to the palace. It takes an hour at a slow pace, and you arrive at the same gate the imperial army arrived at on 3 May 1868 — and you understand why it is still standing.

Scroll to Top