A genealogy of the roof garden — from idea to technique, and on to the department-store roof

The history of a technology is usually told as a single river: someone invents, someone improves, and in time it becomes the standard. The history of the green roof does not run that way. Putting soil on a roof was begun at least three separate times, out of wholly different motives, by people who knew nothing of one another. One began for fire safety, one raised it as an idea, and one did it to draw customers. For a long while the three did not meet; each grew up in its own country. The load calculations and root-barrier standards examined in yesterday's second instalment descend from only one of these three lines — the postwar German one.

Why trace the lineage? Because what reaches anyone setting out to start a rooftop farm today is weighted heavily towards just one strand. Say rooftop greening in Japan and the conversation becomes one about environmental performance, about standards for load and waterproofing. That is the German line. Yet Japan's roofs carry an entirely separate history of their own, more than a century long — a history of amusement parks, of zoos, of beer gardens. The two are, as things stand, barely connected at all. When we speak of rooftop farming as a new undertaking, we are forgetting our own rooftop culture wholesale.

This article arranges the three lineages not by date but by motive. First, nineteenth-century Germany, setting out from fire safety and durability. Then the modern architecture of the 1920s, which raised the roof garden as an idea. Then early-twentieth-century Japan, which invented the roof as commerce. Against that, we set the process by which the three were bound into one technical system in postwar Germany alongside the process by which Japan's roofs closed for quite different reasons. Finally we watch those supposedly closed Japanese roofs open again, in another form, in the 2000s.

Carl Rabitz — born 22 December 1823 in Halle, died 10 April 1891 in Berlin — is usually known in architectural history as the inventor of the Rabitz system: wire mesh stretched beneath timber beams and rendered with plaster to make a fire-resistant ceiling, patented in 1864. His preoccupation throughout was fire. The nineteenth-century city was an accumulation of timber and combustibles, and the roof was the surface most exposed to flying sparks. Rabitz proposed covering that roof with earth and plants. In 1867 he published a pamphlet titled Naturdächer von vulkanischem Cement — Natural Roofs of Volcanic Cement — and in the same year exhibited a greened flat roof at the Paris World Exposition.

The order of the motives is worth noting. For Rabitz the greenery was not the end but the means. A layer of earth and plants protects the roof from sparks, shields the waterproofing from ultraviolet light and temperature swings, and extends the roof's life. Landscape, wildlife and cooling were, if present at all, secondary arguments. Flat-roof greening spread in nineteenth-century Germany for this practical reason rather than out of environmental awareness. The tar and cement layers then used for waterproofing were vulnerable to sun and thermal movement, and laying gravel or soil over them to protect them was, by the standards of the day, an entirely reasonable construction method. Rooftop greening began not as greening but as roof protection.

That starting point still shadows roof design today. As we saw yesterday, the most urgent layers in a green-roof section are the waterproofing and the root barrier, and the plants are merely the last sheet laid on top. It is not that the greenery is primary and the building secondary; the greenery sits on a system built to protect the building. Rabitz's ordering has not been reversed in more than 150 years. The reason any rooftop-farm proposal has to begin from the building's requirements comes back, in the end, to precisely this.

Le Corbusier and Pierre Jeanneret set out the Five Points Towards a New Architecture in a lecture at the Weissenhof Siedlung in Stuttgart in 1927; a revised text appeared in 1929 in the first volume of the Oeuvre complète. The five points are the pilotis, the roof garden, the free plan, the horizontal ribbon window and the free façade. That the roof garden stands second is no small matter. It was presented not as ornament nor as a bonus but as a consequence following directly from structure — a question of what to do with the flat roof that reinforced concrete had made possible.

Le Corbusier's logic ran on two levels. One was practical: unlike a pitched roof, a flat roof is usable surface, and it should be given over systematically to domestic use. A performance argument came with it — that the layer of earth and plants shields the roof slab from temperature change and lends the building a steadier temperature and humidity. The other level was an idea. A building occupies ground; that occupied area should be given back on the roof. The green displaced by the footprint was to be recovered in the vertical dimension, and this was the thought entrusted to the roof garden. The roof terrace of the Villa Savoye (1928–1931) is known as the clearest built statement of it.

What this lineage bequeathed, however, was mainly language rather than technique. The modernist roof garden came with no standards for root resistance or drainage. Many of Le Corbusier's buildings are in fact known to have suffered from roof leaks in later years. The roof garden as an idea passed to architects the world over, and each time the site wrestled with waterproofing on its own. The norm — put a garden on the roof — spread first and alone; the standardisation of the technique that would support it advanced somewhere else entirely, and nearly half a century later.

At the very time modern architecture was discussing the roof garden as an idea, America and Britain were assigning it an entirely different job: drawing crowds. In April 1935 a roof garden of roughly 0.75 acres opened on the eleventh floor of the RCA Building at 30 Rockefeller Plaza in New York, designed by the landscape architect Ralph Hancock with Raymond Hood. Sections modelled on the garden styles of thirteen nations stood side by side; the central International Garden held a meandering stream and two thousand plant varieties; and there were besides a children's section, a modern-style garden and a vegetable patch. Costumed hostesses staffed it, the planting was lit at night, and admission was charged. The roof garden was built not as an idea but as equipment with a plausible return. Its life was short, however: the nation-themed gardens were demolished by 1938.

Trevor Bowen, the managing director of Barkers, saw those gardens and brought the same Hancock to London. What was built between 1936 and 1938 on the roof of the Derry & Toms department store in Kensington became known as the Kensington Roof Gardens. Its area was about 6,000 square metres (65,000 square feet) and it cost £25,000. Divided into a Spanish garden, a Tudor garden and an English woodland, it was planted with over a hundred species of tree, and flamingos were let loose on its pond. When it opened on 9 May 1938 it was reckoned the largest single roof garden in the world. Tree preservation orders followed in 1976, and in 1998 it was placed at Grade II on the register of historic parks and gardens.

Both roofs are instructive structurally as well. The Derry & Toms roof is reported to have been designed to bear one hundredweight per square foot, with drainage beneath. Converted, that is roughly 550 kilograms per square metre — very nearly the middle of the 390–730 kilogram range for intensive greening that we established yesterday. The Rockefeller Center garden was heavy in the same way: its making consumed 3,000 short tons of soil, 100 short tons of rock, 100,000 bricks, 2,000 trees and shrubs, 4,000 small plants and 20,000 bulbs. Divide the soil alone across 0.75 acres and the arithmetic comes to roughly 900 kilograms per square metre — though it was not, of course, laid evenly, so the figure indicates a scale rather than a load.

The era in which roof gardens were built as individual works, one at a time, ended in the 1960s. One of its last large works was the roof garden of the Kaiser Center in Oakland, designed by Theodore Osmundson & Associates and completed around 1960 along with the building. Edgar Kaiser is said to have taken his inspiration from the gardens of Rockefeller Center, and the garden is recorded as the first built in the United States after the Second World War. But that lineage — large, costly, one-off — cannot green a tenth of a city's roofs. To multiply, you need a method that is cheap, light and undemanding, and a standard that yields the same quality whoever builds it. Germany took that on.

Postwar German roof greening rose out of research rather than out of works. In Berlin, Reinhard Bornkamm is reported to have published work on roof vegetation in 1961, and around 1970 Gerda Gollwitzer and Werner Wirsing and others are reported to have advanced research on the components — root-repelling materials, waterproof membranes, growing media, drainage layers. By the late 1970s the thin, light extensive system was established. Greening a roof as a covering rather than as a garden: that shift in conception lowered unit cost and weight together and created the conditions for spread. The thinking behind the lightweight artificial media examined yesterday comes from here.

Standardisation followed. The Landscape Development and Landscaping Research Society (FLL) was founded in 1975; its green-roof guidance is reported to have appeared first in 1982 as Principles of Green Roofing, and from 1992 to have been revised repeatedly as the Guideline for the Planning, Construction and Maintenance of Green Roofing. The 2018 edition mentioned yesterday belongs to this line. The FLL guidelines are not law, but within Germany they are treated as a de facto benchmark, and they have underpinned standards in other countries too. The upshot is that roughly one German roof in ten is now estimated to be greened. Rabitz's fire-safety expedient had become, 115 years later, a document governing an entire industry.

Japan's rooftop culture began neither with an idea nor with fire safety. It began with trade. Its origin is usually given as 1903, when the Shirokiya drapery store in Nihonbashi installed a games room with a seesaw and rocking horses. In 1907 Mitsukoshi, also in Nihonbashi, opened a roof garden billed as a 'garden in the air': on a plot of about 60 tsubo — roughly 198 square metres — stood a fountain, a pond, a wisteria trellis and bonsai, a rotating panorama and a telescope. In the same year of 1907 the Matsuya store in Kanda opened a rooftop viewing ground, and in 1912 Daimaru's Kyoto store put a roller-skating rink and a music hall on its roof.

The installations grew year by year. In 1923 the Matsuzakaya store in Ginza opened a zoo on its roof at the time of its opening, complete with lions and leopards. In November 1931 the Matsuya store in Asakusa opened 'Sports Land', which added to a small zoo and mechanical horses an eight-seat ropeway called the Kokutei, the 'aerial boat'. The roof was no longer somewhere to drop in while shopping; it had become a reason to come at all. Recovery after the war was quick: Matsuya Asakusa reopened Sports Land in 1946 on 150 tsubo of its ground floor, added a twin flying tower on the roof in 1949, and in 1950 introduced the sixty-seat Sky Cruiser.

What matters is that this lineage did not take soil or plants as its subject. Mitsukoshi's 'garden in the air' was a garden, but what followed moved towards rides, animals and restaurants. Where Germany treated the roof as a surface to be protected and modern architecture as ground to be given back, the Japanese department store treated it as simply another floor. The difference still tells. Japanese reluctance about going up onto a roof is comparatively slight, because the roof was for a long time a quasi-public commercial space; and conversely the stock of technical standards for Japanese roofs is thin, because the roof was treated neither as farmland nor as a greening surface.

The heyday of the rooftop amusement park is placed from the mid-1950s to about 1970. Explanations of the decline that followed usually cite the rise of theme parks and video arcades. That is true, but what proved decisive was regulation. After two catastrophes — the Sennichi department store fire of 1972 and the Taiyo department store fire of 1973 — the fire regulations were revised to require that half of a roof be kept available as an evacuation area. With half the roof area unusable, the commercial case for large rides collapses. The rooftop amusement park lost half its floor to the law before customers had tired of it.

The outcome shows plainly in the figures. A survey as of October 2023 found that, of all 170 stores belonging to the Japan Department Stores Association, five still operated a rooftop amusement park: Matsuzakaya Takatsuki, Matsuzakaya Nagoya, Daiwa Korinbo, Iyotetsu Takashimaya and Hamaya. A rooftop culture built over a century has effectively vanished. And what matters is that it did not vanish because the roofs became unusable. The roofs are still there. Only the use was lost.

This history confirms once more a point the series has made since its first day: that what sits on a roof is determined neither by demand nor by technology but by regulation. The opening instalment showed rooftop greening rising after Tokyo's ordinance made the filing of a greening plan compulsory in 2001, and Toronto's mandate reported repealed in October 2025. Here we are looking at the same mechanism running the other way — regulation taking a use away from the roof. The roof is a surface that opens when regulation opens it and closes when regulation closes it.

When Japan's supposedly closed roofs opened again, it was not as amusement parks but as places of production. On 28 March 2006, on the roof of the Kami-Pulp Kaikan in Ginza 3-chome, Chuo Ward, Tokyo, at a height of about 45 metres, the keeping of European honeybees began: the Ginza Honeybee Project. The honey harvest grew from roughly 150 kilograms in 2006 to about 260 in 2007, about 430 in 2008 and about 800 in 2009. Bees are reckoned to range some two kilometres from the hive, which places the Imperial Palace, the Hamarikyu Gardens and Hibiya Park within their compass. A second apiary was established on 8 March 2010 at Marronnier Gate in Ginza 2-chome.

The undertaking did not stop at bees. According to the Tokyo Metropolitan Government's Bureau of Environment, the 2024 season's harvest reached roughly 2.7 tonnes, and the honey is supplied to department stores and restaurants as a local-production, local-consumption product. The work has widened into the promotion of rooftop greening itself, the growing of sweet potatoes and their distilling into shochu, and the cultivation of trees for washi paper; more than 18,000 people are reported to have taken part in its environmental education. It has spread to other regions too, supporting beekeeping in Hiroshima and Okayama, and it received the award of excellence at the first Tokyo NbS Action Awards.

What this case means within a century-long lineage is clear. Japan's roofs were used as sales floors and amusement grounds for far longer than they were ever used as gardens. To those roofs a new use — production — was added in the 2000s. The people doing it were not department stores but volunteers from the neighbourhood, moved not by footfall but by building a relationship with the district. And the honey they gather is delivered to department stores: the produce of the roof returns as merchandise to the very places that once carried rooftop amusement parks. The uses of the roof have been rewritten from commerce to amusement, and then, by way of disappearance, to production. Rooftop farming is not a new invention but a fourth way of using a surface that had fallen vacant.

This article has not told the idea of soil on a roof as one technical history but traced it as three lineages with different motives. The first is the lineage of fire safety and durability. Carl Rabitz (1823–1891) was a master mason who patented the Rabitz system in 1864, published Natural Roofs of Volcanic Cement in 1867, and exhibited a greened flat roof at that year's Paris World Exposition. The second is the lineage of the idea: Le Corbusier and Pierre Jeanneret set out the Five Points Towards a New Architecture in a lecture at the Weissenhof Siedlung in 1927 and placed the roof garden second among them. The third is the lineage of commerce, beginning with Shirokiya's games room in 1903 and Mitsukoshi's 'garden in the air' in 1907. Against these we set the process by which the first lineage hardened into standards in postwar Germany and the process by which the third disappeared in Japan.

Three points deserve restating by name. First, green-roof technique is not a product of environmental thought: its starting point in 1867 was fire safety and the protection of the roof, and the greenery was a means. Second, the idea and the technique spread separately. Le Corbusier's Five Points of 1927 carried no standards, and the standards themselves were assembled in Germany nearly half a century later — the FLL founded in 1975, its guideline reported to have first appeared in 1982, revisions continuing to the 2018 edition, and roughly one German roof in ten now estimated to be greened. Third, what closed Japan's roofs was regulation: after the Sennichi fire of 1972 and the Taiyo fire of 1973, revised fire rules made half of a roof an evacuation area, and as of October 2023 just five of the 170 member stores of the Japan Department Stores Association still ran a rooftop amusement park.

There are also things not known, and reservations to state. For the German research history — Bornkamm's work of 1961, the component research around 1970 by Gollwitzer and others, the 1982 first edition of the FLL guideline — the primary sources could not be traced at the time of writing. These rest on secondary accounts and should be read as carrying possible error in their finer dates. The estimate that about a tenth of German roofs are greened likewise leaves the denominator, the total roof area, undefined, and should be taken as a range. The design load of one hundredweight per square foot at Derry & Toms is a reported figure and not one checked against the period's structural calculations. An objection is fair, too: the arrangement into three lineages is a frame for explanation, and in practice people and information travelled back and forth across borders. Hancock's own passage from Rockefeller Center to London is the counter-evidence. It would be more accurate to say the lineages were not independent so much as never bound together.

As the third instalment of the series The roof as farmland, this article has laid bare the pedigree of the standards and figures handled by yesterday's Designing for weight and water — the physics that turns a roof into farmland. The load calculations and the root-resistance trials alike are products of a lineage that set out from fire safety and was standardised over a century in Germany. That lineage does not, however, explain everything. New York's commercial rooftop farms, the public experiments that began with Chicago City Hall, Basel's biodiversity-driven roofs, Singapore's food-security model — the world holds roofs that use the same technique out of entirely different motives. Tomorrow's fourth instalment crosses those four motives and asks how a difference in motive divides the design. It moves the viewpoint from history to the present, and from Japan to the world.