What is a rooftop farm good for? — between 633 hectares and 1,284

This series has handed out, one at a time, the tools for seeing a roof as farmland. The first instalment counted the area, the second took the physics of weight and water down to the level of design, the third traced the genealogy, the sixth followed the allocation of cost. The tools are assembled. But the better assembled they are, the more conspicuous the question we kept deferring: what, then, is a rooftop farm good for? Both defenders and critics tend to swap that question for another. The defenders say it is good for a city to have more greenery; the critics say you cannot feed a city that way. Neither is wrong, and neither is answering the other. They fail to meet because nobody settled, before the argument began, what the roof was supposed to deliver.

This article splits the expectation into three and tests each against figures. First, food: how much does a rooftop farm actually put on a city's table? Second, environment: is growing it overhead really lighter than trucking it in? Third, green space policy: has the duty to green a roof become an alibi for the greenery lost at ground level? On all three the answer moves towards not as much as advertised. Then we retrieve a fourth use — one nobody was expecting — from a school roof in Rome. Finally, granting every one of the limits, we set out in order the first steps a citizen can take. Counting limits is not the same as giving up. Only once you have settled what a thing cannot do can you settle what it can.

One thing should be said in advance. Of the seven days planned, this series could not publish its fourth instalment, on cases around the world, or its fifth, on where Japan stands. This article therefore does not sum up seven pieces of argument; it stands on the five that were actually published — area, physics, genealogy, cost, and this one. What those two missing instalments would have covered, above all the present state of Japan's roofs, is filled in here only as far as figures we could verify allow, and that is no substitute. Having chosen not to rewrite the shape of the series after the fact, we record the gap as a gap.

Since 2000 Japan's Ministry of Land, Infrastructure, Transport and Tourism has run an annual survey of rooftop and wall greening work nationwide. Its latest results, published on 12 December 2025, put newly constructed green roofs in 2024 at about 14.5 hectares and green walls at about 4.6. Cumulatively, from 2000 to 2024, that is roughly 633 hectares of roof and 131 of wall. The first instalment of this series cited the then-current figure of about 615 hectares through 2023; adding another year does not change the character of the number. Twenty-four years have accumulated 633 hectares, and the most recent year added 14.5 — the growth, as the first instalment showed, has already levelled off.

Now set those 633 hectares beside the ground. According to the agriculture ministry's account of Japan's allotment gardens, fiscal 2024 counted 4,273 gardens nationwide, 188,713 plots and 1,284 hectares. By operator, 1,977 were opened by local authorities, 1,433 by farmers, 449 by companies and non-profits, and 414 by agricultural cooperatives. Divide 1,284 hectares by 188,713 plots and each plot comes to about 68 square metres — that much soil in the hands of each of more than 180,000 plot-holders. In fiscal 1992 the same figures stood at 691 gardens, 56,727 plots and 202 hectares, so the area has multiplied roughly six and a half times in a little over thirty years.

Side by side, two things emerge. The gardens on the ground hold about twice the area of twenty-four years of rooftop greening; and, decisively, all 1,284 hectares of it is soil meant to be eaten from. Of the 633 hectares overhead, most is thin-layer greening of sedum and the like, and how much carries edible crops these statistics do not say. The comparison has limits, of course. Allotment gardens are a nationwide figure that includes flat suburban land, while a green roof can exist in dense central districts where nothing else can — a property no area figure captures. Even so, if the object is simply to multiply the places in a city where a person can touch soil, these two numbers have very nearly answered which route is efficient.