Gardens Above The Street

AThe idea that a layer of living plants on a roof will cool the building beneath it, and the street outside it, has passed so quickly from novelty to orthodoxy that the mechanism behind it is seldom examined. Municipal grant schemes across northern Europe, of which there were fewer than a dozen in 2005 and more than two hundred by 2021, almost without exception measure their success in a single currency: square metres of roof brought under vegetation. Embedded in that metric is an assumption about where the cooling comes from. Leaves intercept the sunlight that would otherwise strike the waterproof membrane, the argument runs, and the roof consequently stops behaving like a hotplate. It is an appealing account. Whether it is the correct one became answerable only when sensors cheap enough to be abandoned outdoors for years at a time started to be installed in quantity.

BThe sensors have told a more complicated story. Between 2013 and 2019 a team led by Marta Kowalczyk at the Institute of Urban Fabric in Ghent logged surface and air temperatures on forty-six planted roofs, choosing buildings that differed widely in what grew on them but shared a growing medium some ten centimetres deep. Kowalczyk's finding, released in stages, was that the cooling measured above a roof tracked the water held in its growing medium far more faithfully than it tracked any property of the vegetation: roofs whose medium was close to saturation cooled the air above them by an average of 2.9 degrees Celsius, whereas roofs that had dried out managed 0.4, a gap several times larger than any produced by exchanging one planting scheme for another. What a roof is growing, she contends, matters chiefly because some species hold moisture in the medium for longer; the leaves are a means of storing water, not a parasol.

CNot everybody accepts so austere a reduction. Tomas Duvalier, a building physicist at Kestrel College, maintains that Kowalczyk has mistaken a local condition for a general law, since Ghent and the other cities in her sample receive summer rain at intervals short enough that a roof is seldom dry for long, and a variable that never runs out will always look decisive. In continental interiors, he argues, the medium is exhausted within days, and it is there, and only there, that the architecture of the canopy, its height, its density, the angle at which the leaves are held, begins to do the work. Duvalier's own evidence, however, rests on six roofs followed through a single fourteen-month season, and two of those were test rigs on a college campus rather than occupied buildings. He concedes that the sample is thin. What he refuses to concede is that his objection is thin with it, since a small sample drawn from the right climate may still expose a mechanism that a large one drawn from the wrong climate can never reveal.

DThe quarrel is not merely academic, because the second service claimed for planted roofs draws on the same water. A roof detains rain only if its medium has room left to absorb it, so the saturation that produces the sharpest cooling is also the condition in which retention is weakest. Kowalczyk's own dataset makes the tension plain: across the storms she catalogued, 19 per cent of the rain falling on saturated roofs was held back, against 19 of the 42 storms in which a dried-out roof absorbed everything that reached it. A design that maximises one service will therefore degrade the other, and although greater depth narrows the conflict, no depth escapes it altogether. Cities that want both must decide which they need in which month. Drainage engineers have understood the point for decades, which is why the tanks they design to hold storm water are deliberately emptied between downpours, whatever that regime may cost anything growing on top of them.

ESubsidy design has not caught up with any of this. Because grants are almost everywhere paid by the square metre planted, a developer who lays the thinnest permissible medium across the widest available area collects the largest payment while delivering the least of either service, a practice common enough to have earned a nickname among contractors. The revision adopted by Rijnstad in 2018, which pays according to the litres of runoff a roof detains during a standard test storm rather than the area it covers, remains unusual. It is also dearer to administer, since every roof must be gauged rather than merely photographed. Officials there report that the average depth of medium on new roofs climbed from eight to fourteen centimetres inside three years, while the total area of new planting fell by roughly a fifth. Two smaller municipalities have since copied the formula, and both report the same shift in what developers put forward for approval.

FWhere does this leave the roof itself? The honest answer is that its performance is contingent in a way the early advocacy never admitted: a construction that cools a street through a maritime summer may do very little in a dry August inland, and no roof can be relied upon for cooling and for flood control in the same hour. Kowalczyk's account is the better description of temperate northern Europe, and the weight of the monitoring evidence supports it there. Duvalier is probably right that his mechanism takes over once moisture becomes scarce, though the evidence he rests it on is too thin to show how quickly the changeover occurs. Neither position is likely to be abandoned. What deserves to be abandoned is the notion that a roof can be judged by looking at it.