Designing for weight and water — the physics that turns a roof into farmland

On the ground, design can begin with the crop. Decide what you want to grow, build soil to suit it, bring in water if you must. There is rarely any need to ask what lies beneath the soil. On a roof that order is reversed. The first question is how many kilograms this slab will permit; the second is how the waterproofing will be protected; the third is where water will enter and where it will leave. The choice of crop comes last of all. Yesterday's opening instalment described the roof as farmland stripped at the outset of the most basic freedom of all, depth of soil. Today we open up the contents of that constraint, following the numbers and the order of the layers.

What happens when that order is broken is quite clear. Pile the soil too deep and the structure will not hold. Tear the waterproofing at a single point and the floor below gets wet. Block the drains and water ponds on the roof, and that water becomes weight in its turn. Leave out irrigation and the whole thing dies within days of midsummer. The failures of rooftop farms arrive as failures of equipment rather than failures of agriculture. This article is therefore written as a piece about design, not as a growing guide. Its subject is four physics — weight, water, wind and heat — and every one of them is tied to the others.

The way they connect often runs against intuition. Weight, for instance, is the single greatest constraint on a rooftop farm, and at the same time almost the only defence it has against wind: the lighter the medium, the more readily it is carried away. Water is the lifeline of the plants and the gravest threat to the building, and soil that has drunk its fill is soil at its heaviest. What produces the cooling is not the soil itself but the water the soil is holding. Lighten one thing and something else grows weaker — designing a roof is nothing other than deciding where to draw the line in that tug of war.

In Japan, what a roof may carry is set out first by the table of imposed loads in Article 85 of the Building Standard Law Enforcement Order. Row (8) of that table names roof plazas and balconies and directs that their values follow row (1) — residential rooms, and bedrooms or sickrooms in non-residential buildings. The row (1) figures are 1,800 newtons per square metre when calculating the floor structure, 1,300 newtons for girders, columns or foundations, and 600 newtons when calculating seismic force. For buildings used as schools or department stores, however, the values of row (4), the sales floors of department stores and shops, apply instead: 2,900, 2,400 and 1,300 newtons.

Divided by gravitational acceleration, 1,800 newtons comes to roughly 184 kilogram-force. Up to 180 kilograms per square metre sounds generous when put that way, but the meaning of the number must not be mistaken. An imposed load is a statistical allowance for loads that move and change — people, furniture, stock. The soil, drainage material and trees of a green roof do not properly belong there at all. Saitama Prefecture's published material on rooftop greening likewise states that rooftop greening is limited by the building's imposed load and refers to the Enforcement Order. The allowance does indeed constrain what is possible; yet the soil itself belongs to a different category of load.

For a new building the matter is simple: fold the weight of the greening into the design conditions from the outset and, if need be, add beams and columns. The difficulty lies with existing roofs. How much margin was assumed at design time, and how much of it may legitimately be used, cannot be known without consulting the drawings and the structural calculation documents. One cannot say there is a roof, so it can be a field; one should say there are structural calculations, so it can be considered. When citizens set out to start a rooftop farm, the first obstacle is neither soil nor plants but a single set of documents.

Structural design divides loads broadly into dead loads and imposed loads. A dead load is the weight of the building itself — the slab, the finishes, the services, everything that does not move. An imposed load is a statistical allowance for what does move: people, furniture, stock. The layers of a green roof, from the ballast over the waterproofing through the drainage material and growing medium to the plants themselves, belong in principle to the former. It is because of this distinction that a greening plan is required to proceed alongside the structural design from the very start. Putting soil on a roof is not redecoration; it is an act that increases the building's own weight.

Line up the numbers and the situation becomes plain. Thin, light extensive greening is put at roughly 50–120 kilograms per square metre; intensive systems with deep soil at 390–730. The lower bound of the intensive range, 390 kilograms, exceeds even the imposed load allowed for an office — 2,900 newtons, about 296 kilogram-force. An intensive roof, in other words, is not the kind of thing that can be added later out of spare capacity. Retrofits on existing roofs drift towards extensive systems not because the philosophy of greening has changed, but because arithmetic compels it.

A second point matters just as much: in what state the weight is measured. Soil dry and soil that has drunk its fill of rain are entirely different weights. North America has standards for this — ASTM E2397, which sets out the determination of dead and live loads for vegetative roof systems, and ASTM E2399, which measures maximum media density for dead-load analysis — and both are reported to include procedures that saturate the medium before density is taken. The weight to design against is not the weight on a fine day. It is the weight at the heaviest moment, after heavy rain, when the soil has taken up all the water it can hold. A calculation that misses this misses the most dangerous state of all.

A green roof in section is, from the bottom up, a waterproof layer, a root barrier, a drainage layer, a filter, the growing medium, the plants, and often irrigation pipework. Five or six layers appear here that a field on the ground does not have at all. Think of a rooftop farm as a field whose soil was carried upstairs and the meaning of that stack disappears. More accurately it is a composite: a second system, built to keep plants alive, stacked on top of a first system built to keep the building dry. Neither side's convenience can govern alone, and the points where their demands collide — the position of the drains, the detailing at upstands — are precisely where the design is won or lost.

The most urgent of these is the root barrier. Roots — above all species that send out rhizomes — find the seams in the waterproofing and the weak points at upstands, and go through them. Europe has EN 13948, a test standard for the root resistance of waterproofing sheets, and the German FLL guidelines (2018 edition) are reported to be stricter still, imposing a two-year trial using firethorn and couch grass and a four-year trial using grey alder and couch grass. Plants are set to attack for years, and only if the membrane survives is it passed. Root resistance, in other words, is less a materials test than a form of cultivation trial. In Japan the common practice is to lay a separate root-barrier sheet over the waterproofing.

Drainage and filter layers are unglamorous, and these two decide how long a rooftop farm lasts. The drainage layer carries water to the outlets without letting it stand, while holding a certain quantity in reserve. The filter stops fine particles from washing down and clogging it. If the drains block, the roof becomes a shallow pond, and ten centimetres of standing water is by itself an added load of roughly 100 kilograms per square metre. The margin examined in the previous section is eaten away exactly like this, quietly. The practitioners' rule that the first thing to inspect on a rooftop farm is not the crop but the drain rests on that piece of arithmetic.

The soil on a roof differs from field soil in its very materials. The lead role goes to fired, expanded aggregates — grains of clay, shale or slate baked at high temperature until they foam — and to naturally porous volcanic gravels such as pumice and scoria. Expanded clay aggregate (LECA) is fired in a rotary kiln at around 1,200°C, where trapped gases expand and form thousands of small bubbles, leaving the grain porous. Its bulk density runs at roughly 250–510 kilograms per cubic metre depending on grade. Given that mineral soils are generally put at about 1.1–1.6 grams per cubic centimetre — that is, 1,100 to 1,600 kilograms per cubic metre — this is a third to a fifth of the weight.

Why, then, can something so light hold water? The key is where the pores are. Water is retained not only between grains, in interparticle pores, but inside the grains themselves, in intraparticle pores. Precisely because water can be held within the grain, the space between grains can be left free for drainage. That seemingly contradictory property — it drains fast and still holds — comes from this double-decked pore structure. In an experiment on the roof of National Central University in Taiwan, a 10-centimetre lightweight artificial substrate was reported with a field capacity of 0.45 by volume, porosity of about 0.46, and saturated hydraulic conductivity of 339.8 millimetres per hour. Conductivity of 340 mm an hour passes a downpour straight through, and the same material still holds water to 45 per cent of its volume.

Lightweight media are not, however, a cure-all. Mixing in organic matter raises water retention and fertility, but organic matter decomposes and diminishes, so the substrate loses depth over a few years. A light grain is also a grain that blows away. And the most fundamental limit is that depth itself is bound by load. A study comparing substrates of 10 and 30 centimetres confirmed that deeper substrates retain more water. Everyone knows that going deeper would help, and no one can go deeper. The soil on a roof sits in a place where law and structure forbid the answer that is already known.

Wind is stronger on a roof than at ground level, and it does not blow evenly. Airflow separates at the building's edge, so uplift at the parapet and the corners is far greater than over the middle of the roof, and scour strips the medium away at the same time. When a rooftop farm fails, it rarely fails in the middle; it usually fails at the edge. North America has ANSI/SPRI RP-14, a wind design standard for vegetative roofing systems, reported to provide a method for deriving allowable wind speeds from building height, parapet height and exposure category. That it is built on the standard for ballasted single-ply roofs says a great deal about its character.

There are two essentials in response. The first is a band around the roof edge and around anything protruding from it where no planting is placed — a vegetation-free zone — held down with gravel or similar. Its width is generally put at roughly 50–90 centimetres, and practice guidance also calls for 300 to 500 millimetres of gravel around upstands. The band doubles as a firebreak. The second is to use the medium and gravel themselves as ballast. Most vegetative roofs are neither adhered nor mechanically fastened; they resist wind by the weight sitting on them and nothing else.

Here the meaning of weight inverts. As the earlier sections showed, weight is the greatest constraint on a rooftop farm. Against wind, that same weight is almost its only weapon. Push lightweighting to its limit and the risk of the medium blowing away rises; add weight for wind resistance and the structural margin is consumed. There is no simple optimum of the lighter the better in roof design because these two demands collide head-on. Design is the judgement of where to draw the line through that collision, and the line falls in a different place for every building height and every site.

Temperature is the first benefit anyone cites for a green roof. The US Environmental Protection Agency reports that the surface of a green roof can be up to 56°F — roughly 31°C — cooler than a conventional roof, that nearby air temperatures can fall by up to 20°F (about 11°C), and that in one case the cooling load was cut by 70 per cent and indoor air temperature lowered by 27°F (about 15°C). A 2023 meta-analysis reported summer roof surface temperatures lower by an average of around 30°C, while measurements at Chicago City Hall put its roof 1.4–4.4°C cooler than conventional roofs. The spread in these figures is only to be expected: the climates, the construction and the plants all differ.

What matters is where that cooling comes from. In the experiment on the roof of National Central University in Taiwan, a plot of native species in a 10-centimetre lightweight artificial substrate lowered indoor temperature by 5–6°C on average through the summer, and by as much as 10.65°C, relative to a control roof. The largest contributor to that reduction was neither insulation nor shading but evapotranspiration, followed by substrate moisture, air temperature and vapour pressure. What cools the building, in other words, is not the layer of soil but the latent heat carried off as the water held in that soil evaporates. It is closer to the truth to think of a green roof not as insulation but as a cooling device that runs on water as its fuel.

That sentence carries a heavy operational implication. A dry green roof does not deliver the performance expected of it. Let substrate moisture fall and evapotranspiration stops, and the cooling stops with it. Irrigation is care that keeps the plants alive and, at the same time, the operation of equipment that keeps the roof performing. The same holds for rainwater. A study in Opole, Poland, reported that green roofs retained roughly 44 per cent of heavy rainfall and retained completely any daily rainfall of 5 millimetres or less, on both extensive and intensive roofs. Yet with rain continuing three days or more the reduction fell by 20 per cent, and after a dry spell of three days or more it rose by 20 per cent. A roof's performance is set not by the depth of its soil alone but by how dry that soil happens to be right now.

This article has been carried by a single principle: read a rooftop farm's drawings from the bottom up. What we placed first was Article 85 of the Building Standard Law Enforcement Order. Roof plazas and balconies are treated like residential rooms, at 1,800 newtons per square metre for the floor calculation, 1,300 for girders, columns and foundations, and 600 for seismic force — and at 2,900, 2,400 and 1,300 respectively for schools and department stores. We then established that these numbers are not numbers for soil, and recounted the weight of greening as dead load. From there we descended through the section of layers, into the materials of the medium, out into the wind, and on to heat, arriving at the conclusion that what cools a roof is not the soil but the water the soil holds. The weight we began with and the water we ended with are joined at a single point: saturated soil.

Three points are worth carrying away. First, roof design is a question of dead load, not imposed load: extensive systems run at roughly 50–120 kilograms per square metre and intensive ones at 390–730, and even the lower bound of the intensive range exceeds an office's imposed load of 2,900 newtons, about 296 kilogram-force. Second, artificial lightweight media are light and still hold water because of a double-decked structure of intraparticle and interparticle pores; expanded clay fired at around 1,200°C has a bulk density of roughly 250–510 kilograms per cubic metre, a third to a fifth of mineral soil. Third, the cooling of a roof is carried by evapotranspiration: in the National Central University experiment in Taiwan a 10-centimetre substrate lowered indoor temperature by 5–6°C on average in summer and by up to 10.65°C, with evapotranspiration the largest contributing factor.

There are also things not known, and reservations to state. The ranges of weight and the temperature reductions cited here are gathered from cases differing in climate, construction, planting and standard of maintenance, and none of them transfers unchanged to any given roof. The EPA figures are upper values observed under particular conditions, not averages. The Taiwanese and Polish results were obtained in subtropical and central European climates, with no guarantee that they carry over to a Japanese rainy season and a Japanese heatwave. For the ASTM, FLL and ANSI/SPRI standards, the full texts could not be obtained at the time of writing, so the descriptions here are confined to their names and purpose and no detailed figures are quoted from them. And the weightiest objection to anticipate is this: reduce a roof too far to a problem of engineering and the labour and the pleasure of the people working on it, and the taste of what grows there, drop out of view. Design is a necessary condition, not a sufficient one.

As the second instalment of the series The roof as farmland, this article has translated into the language of design what yesterday's opening instalment presented as area. The greenery that sits on a roof is, before it is the product of any idea, the consequence of load, waterproofing, drainage and irrigation. That body of technique did not, however, fall from the sky. Who first thought of putting soil on a roof, when, and out of what motive? Why was a root-resistance standard demanding years-long cultivation trials assembled in Germany of all places? Tomorrow's third instalment traces how modern architecture raised the roof garden as an idea, how postwar Germany turned roof greening into a technical system, and how Japan arrived at the distinctive culture of the department-store rooftop. We shall be looking, from the side of history, at where today's numbers came from.