Kumamoto Castle: Putting the Stones Back

After the 2016 earthquakes, collapsed walls blocked a passage near Kumamoto Castle's Hohoate Gate. The route had been more than six metres wide. Now the stonework on both sides lay across it. Before large vehicles could reach the keep for repairs, the fallen material had to be recorded, recovered and made safe. A temporary construction route was finally established in March 2017.^[1]

Repairing the tower depended on work somewhere else. That practical dependency is a useful entrance to Kumamoto's history. A castle is an arrangement of buildings, retaining walls, fill and routes, assembled and altered at different times. When those relationships fail, recovering the most recognizable building is only one part of recovering the place.

At Kumamoto, even the instruction to put a stone back raises a historical question: back into which wall, in which position, and with what evidence that it can safely remain there?

Katō Kiyomasa developed the great castle on the plateau called Chausuyama, with 1607 conventionally marking completion. The Hosokawa family succeeded the Katō in 1632 and remained through the rest of the domain era. Kumamoto's history between construction and modern disaster therefore contains generations of occupation, maintenance and changes of use, not simply the long survival of one builder's finished work.^[2]

The defenses organized a demanding landscape. Stone retaining walls supported raised compounds, with gates and passages controlling movement between them. At Iidamaru, south of the central enclosure, the site's interpretation identifies a defensive position connected with a kitchen well and an ammunition store.^[3] Defense depended on usable space and provisions as well as on obstacles presented to an attacker.

When the domains were abolished in the nineteenth century, the castle did not immediately become an inert monument. Its grounds entered army use, and the garrison headquarters moved into the main enclosure in 1874. During the Satsuma Rebellion of 1877, government troops held Kumamoto against the insurgent Satsuma forces for more than fifty days. The keep and palace had burned immediately before the fighting; the cause of the fire remains unsettled.^[2]

The garrison's resistance after that fire makes the distinction between a castle and its tower especially concrete. Losing an imposing building did not eliminate the defensive ground, the enclosures or the protected positions available within them. A modern army could still use a site made for an earlier military order.

The war also changed the material inherited by later repairers. Fire had affected stone as well as timber. When an earthquake struck in 1889, an army report linked some of the destruction to stone weakened in the 1877 fire and ground softened by prolonged rain. These were the investigators' explanations, not a modern laboratory diagnosis.^[4]

A wall damaged in 2016 might consequently contain the effects of more than one disaster. Its visible shape could preserve an old design while its individual stones and joints carried the consequences of fire, earthquake and previous repair. The recovery team had to work with that accumulated history rather than an imaginary castle untouched since Kiyomasa.

After the 1889 earthquake, the army initially dealt with dangerous masonry by bringing unstable sections down. It then undertook a longer repair campaign. A contemporary requirement to reuse the existing stones presented contractors with a difficulty: the castle's large, irregular blocks were unlike the more manageable standardized stones common in their current work.^[4]

The city research center's account describes differences in the resulting repairs. Some walls were restored at a reduced height; elsewhere, stonework was removed to ease passage. Later work could continue the castle's use without reproducing every earlier dimension. The repair itself became a historical intervention, leaving material that the next generation of investigators would need to recognize.

That changes what “old stonework” means. A wall can incorporate early stones in a later arrangement. A repair may preserve a building's military usefulness while altering the way people move around it. A date attached to the first castle cannot settle the history of such a wall, because the arrangement of its parts may belong to another campaign of work.

This is particularly important when interpreting damage. A section that failed after an earlier repair does not prove that all later workmanship was inferior. Its ground, height, loading and previous damage also form part of the problem. The surviving fabric has to be examined as a particular construction, not made into a contest between a famous founder and anonymous successors.

For Kumamoto's present conservators, earlier rebuilding is also something they can encounter from within. Archaeological investigation accompanying dismantling records joints, packing and traces of repair as courses are removed. The work can expose how a seemingly continuous face was put together at different times.^[5]

The investigator is therefore studying a structure that others have already studied, altered and returned to service. What looks like a straightforward task of reconstruction can become the recovery of several previous solutions to the same problem: how to keep this raised ground, and the buildings upon it, in use.

The post-2016 recovery policy supplied a starting point: normally restore the walls to their condition immediately before the earthquake, rather than use disaster repair to recreate an ideal Edo-period appearance. Dismantling was to be minimized and traditional methods preferred. The policy also allowed exceptions where the pre-earthquake condition was already unsafe or where modern reinforcement was necessary.^[6]

This was a consequential choice of reference date. Returning everything to the first building campaign would erase later repairs that had helped the castle survive. Simply reproducing every pre-earthquake bulge, however, might restore a danger along with the appearance. The policy separated the historical condition being sought from the obligation to assess whether that condition was acceptable to rebuild.

The evidence for the desired shape was uneven. Some photographs looked obliquely across the masonry; trees concealed parts of other views. Detailed measurements of every wall before collapse did not exist. Uncollapsed portions could help establish a profile, but a photograph showing where a face stone belonged might not reveal the complete shape of the wall behind it.^[6]

There are therefore two different identification problems. One concerns an object: which recovered stone is this? The other concerns an arrangement: how did the wall run through space, and how were its parts supported? An answer to the first cannot automatically supply the second.

This is also why a newly quarried replacement, even one of similar size, cannot always do the same historical work as the recovered block. The old stone may retain an identifiable position and evidence of previous cutting or use. The aim of recovering it is not simply to obtain enough masonry. It is to preserve as much as possible of the particular construction that stood here.

The requirement begins before the lifting equipment clears the ground. Fallen material still has a location, an order of overlap and a relation to the wall from which it came. Once removed without a record, that part of its history would be much harder to recover.

At Hohoate Gate, the recovery described by municipal archaeologist Kamura Tetsuya began with photographs, individual numbers and surveyed coordinates. Only then were stones collected. Stabilization of the remaining slopes and masonry followed, before the temporary route for construction vehicles was installed.^[1] The sequence made emergency access compatible with retaining information about the collapse.

The city's archaeological account describes survey and lifting as repeated operations, rather than a single photograph followed by wholesale clearance. As one layer of fallen material was removed, another became available to record. Individual stones were measured, photographed on all faces, examined for working marks and assessed for reuse.^[5]

The record of where a stone landed could help the search for its former position. Combined with matches to photographs, it could also help reconstruct how a wall fell. That is a different use of the same information: one investigation seeks the earlier arrangement, another asks how it came apart. Both bear on the eventual rebuilding.

Nor did research stop at the visible blocks. Behind a stone face lie smaller packing stones, and beneath the wall is ground with its own construction history. The city's investigations included geological work as well as excavation. Understanding the supports required attention to material that a finished exterior would hide again.^[5]

The buildings produced related problems. At the collapsed Higashi Jūhakken turret, timber, tiles, plaster and stone had become mixed together. Recovery teams separated and cleaned components, identified their roles and stored reusable material under protection. A city account illustrates the same careful storage of timbers from the Uto turret's annex.^[7]

Here, conservation meant temporarily taking a building farther from its familiar appearance. A numbered timber in storage no longer presents a roof or a room. Yet identifying and protecting that timber can preserve the possibility of returning part of the actual building, rather than replacing it with a newly made likeness. The orderly store and the damaged site belong to one recovery process, even when neither resembles the castle people remember.

Matching thousands of blocks by eye was an obvious place to seek assistance. It was also harder than asking software to find two identical photographs. A stone photographed high on a wall and the same stone lying in a storage yard could present different angles, shadows and visible surfaces. Many neighboring stones had similar outlines.

Gou Koutaki's research at Kumamoto University approached the problem through both two-dimensional contours and three-dimensional surface shape. His report notes the usefulness to craftspeople of distinctive marks such as yaana, the holes made for wedges used in splitting stone. The system combined kinds of resemblance and offered candidates for a user to inspect, rather than treating a similar outline as sufficient identification.^[8]

That combination is intelligible without the mathematics. An outline narrows a search, while an irregularity on a face can distinguish one candidate from another. Changing the camera's position changes the picture; it does not necessarily remove the particular relationships among features on the stone. The task is to recognize an object across that change of view.

In 2022, a joint project involving Kumamoto University, Shinshu University and Toppan reported matching 337 of 370 fallen stones from the Iidamaru five-storey turret's walls—91 percent of that test set. The university stated that the results were being used in the city's repair design.^[9]

The scale of the result matters. It was a demonstrated identification exercise for a particular group of stones, not a report that the software had rebuilt the whole castle. Koutaki's research places measurement and matching between recovery of the blocks and the preparation of the record used for construction. The system assisted those stages; it did not lift, position or certify the finished masonry.^[8]

The distinction leaves room for the craft of rebuilding without making technological help incidental. A proposed match can save an extended search. The blocks must still fit together physically and participate in a stable structure. At Kumamoto, the digital record is useful because it can be taken back to the stones, not because it replaces them.

A wall can remain standing and still demand major intervention. The masonry beneath the Uto turret's annex had bulged outward after the earthquake. In an account published in April 2026, a city research official described why a solution intended to leave the face in place proved problematic.^[10]

Excavation showed that about four metres of the wall lay buried below the later ground level. Adding a heavy supporting structure at its foot would therefore place new weight over historic masonry. It would also change the appearance of the tall wall. After investigating its history and condition, the team chose limited dismantling and rebuilding rather than that form of support.

The difficulty was not a simple choice between keeping something old and replacing it with something stronger. Both alternatives would intervene in the historic structure. Leaving the visible face untouched could still affect the buried part through the added load. The archaeology changed the decision by showing that the apparent foot of the wall was not its actual beginning.

That is a particularly local lesson in why inspection of an exterior is insufficient. A design based only on the wall above ground would address an incomplete object. Recovering its full height and earlier levels made it possible to ask what a proposed remedy would do to material out of sight.

The same official account describes another occasion when more than fifty stones had to be taken out and relaid after an arrangement proved unstable during work.^[10] A position proposed on paper still had to be tested in the building process. Matching an individual block correctly and recovering a satisfactory relationship among many blocks were separate tasks.

Dismantling consequently has an ambiguous place in conservation. It interrupts the arrangement being preserved, yet may be necessary to retain the material and restore a support that can function. Recording each stage does not make that interruption disappear. It makes the intervention accountable to the evidence uncovered, and leaves a record for whoever must repair the wall next.

The Uto turret and the main keep pose different questions. Uto preserves a premodern timber building, though one altered and repaired through its history. The main keep is the steel-framed reinforced-concrete reconstruction completed in 1960. Repair after the 2016 earthquake did not transform that modern structure into Kiyomasa's surviving tower.^[3]

Iidamaru adds a third date to the familiar skyline. Its five-storey timber turret was reconstructed in 2005. When the supporting masonry partly collapsed in 2016, the building remained precariously carried at a corner. Its relatively recent age did not make the wall beneath it, or the consequences of that wall's failure, a purely modern problem.^[3]

These differences determine what can be recovered. At a historic timber building, a surviving post or beam is material from the building's own history. The city's stated approach was to recover and reuse serviceable components, with structural reinforcement considered where necessary. That commitment is different from producing entirely new timberwork to represent a building that has disappeared.^[7]

It is different again from repairing a concrete museum building. The keep's reopening in 2021 returned an important public interior to use, but it could not stand as a certificate that all surrounding walls and turrets were restored. Their materials, dependencies and research problems remained distinct.^[11]

Kumamoto's recognizable outline is therefore a poor unit for measuring recovery. It can look substantially restored while major work remains elsewhere. Conversely, a dismantled historic turret can look absent while its identified components are being conserved. Appearance alone can misrepresent progress in either direction.

The useful question is what operation has actually been completed: recovering fallen blocks, repairing the supporting wall, conserving a frame, reconstructing a lost building or reopening a museum. Each is a real achievement, but each answers a different need. Keeping their names and dates separate allows the site to be understood without assigning the whole castle one misleading label of “original” or “rebuilt.”

A municipal progress report makes this uneven rhythm explicit. It records completion of Iidamaru's stone-wall rebuilding in January 2024, followed by a separate campaign on the turret. The Hirayagura wall was completed in August 2025; Uto's dismantling-and-conservation phase concluded the following month. These were milestones in different operations, not interchangeable declarations that their entire settings were finished.^[11]

Recovery also remains exposed to new events. After another earthquake in 2026, an official August notice announced reopening following safety checks. That notice concerned access at that moment, not the disappearance of the castle's unfinished conservation work.^[12]

What joins the projects is a method of returning from an overall image to particular relationships. A fallen stone needs a former position; the reconstructed face needs support; the support needs ground whose history is understood. A building needs a route by which its repair can be reached. The blocked passage at Hohoate was part of the keep's recovery for exactly that reason.

Putting the stones back is thus both a physical task and an investigation. The aim is a castle that can be used and understood again, while retaining evidence of the work through which it reached the present. The finished view matters. So does the record of what had to be found, retained and changed to make that view possible.