The AI Journal

By 1944, rubble and ash had replaced perhaps a hundred acres of the City of London, and naturalists were doing something that must have looked strange against the skyline: cataloguing wildflowers. Richard Fitter, compiling the record that became London's Natural History, found more than a hundred species of flowering plant growing on the bombed sites of a district that had supported barely twenty before the war (Fitter, 1945). Rosebay willowherb, a plant so rare in London before 1940 that herbarium specimens were prized, became so common on the ruins that Fitter called it the pioneer coloniser of the bombed sites; the elephant hawk-moth arrived behind it, feeding on leaves that had nowhere to grow a few years earlier. Nobody planted any of this. The flora appeared because the ground had been cleared, left alone, and disturbed again by frost, rain and foot traffic, and because nobody was paid to weed it.

The bombsites were cleared decades ago, but the pattern they revealed has not gone away. Oliver Gilbert’s foundational study of British urban ecology took demolition sites, railway sidings, canal banks and derelict yards as seriously as any nature reserve, on the grounds that this is where a city’s wildlife actually concentrates (Gilbert, 1989). The claim keeps being confirmed. A review of thirty-seven European studies on urban wasteland found that abandoned ground with spontaneous vegetation harboured more species, in most comparisons, than the deliberately planted green space around it (Bonthoux et al., 2014). A Canadian study comparing spontaneous vegetation with mown lawn and managed urban forest found the unmanaged ground scored highest on habitat provision, by a wide margin, even though it did worse than lawn on other services such as carbon storage (Robinson & Lundholm, 2012). The wasteland is not a lesser version of a park. On several ecological measures it is a better one, and the species using it are frequently not found anywhere else in the city.

Much of this ground stays open long enough to matter for an unglamorous reason: contamination. A former gasworks, railway yard or scrapyard is frequently too costly to clean to residential standard without a subsidy, so it sits fenced and untouched for years while cheaper sites nearby get built first. The same economics that make a site useless to a developer are what give disturbance the time to do its work, and the same remediation that makes the ground safe to build on, stripping and capping contaminated soil, is usually also what erases the particular, nutrient-poor, structurally varied substrate the rarer specialists depend on. The land is not idle because nobody wants it. It is idle because building on it is expensive, and the expense buys the ecology its interval.

The reason is not neglect in the vague sense but a specific structural feature neglect produces: instability. A managed park is deliberately held at one successional stage, usually an early one, with bare soil and tall vegetation both removed as soon as they appear. A patch of genuine wilderness, left alone for centuries, eventually settles into a different single stage, usually a late one dominated by a few large species that outcompete the rest. Neither extreme is hospitable to the specialists that depend on disturbance itself: plants that need open ground to germinate but cannot survive once grass closes over it, insects that need bare, sun-warmed soil for burrows, birds that nest in the gaps between scrub and clearing. A vacant lot gives these specialists what a park’s constant tidiness and a forest’s constant permanence both withhold, which is unpredictability over a timescale of years rather than decades. The biodiversity is a byproduct of the churn, not of the dereliction as such.

That the mechanism is churn rather than neglect has a counterintuitive consequence: leaving a single site alone forever is not actually the optimal policy, because a site frozen in place eventually loses the instability that made it valuable. A simulation study of business-park brownfields in Germany tested this directly, varying how long sites were left fallow before redevelopment and how much open land was available across a city at any one time (Kattwinkel et al., 2011). The highest conservation value came not from permanent protection but from rotation: setting sites aside for roughly fifteen years, then redeveloping them while new vacancies opened elsewhere, so that the city always held a shifting mosaic of ground at different stages rather than a fixed set of reserves. A landscape-scale study across Britain, using national monitoring records tied to the locations of former landfills, found the same instability inside a single kind of site: species richness for birds and insects was highest soon after a landfill closed and declined as the site aged, while plant richness kept climbing over the same period (Macgregor et al., 2022). Even the sites doing the most ecological work are not doing it statically. They are passing through a stage, and the stage, not the site, is what needs to exist somewhere in the city at any given time.

The obvious objection is that this entire argument asks cities to treat idle land as an amenity at exactly the moment many of them are desperate for housing. Vacant and derelict land is not scenery; in most advanced economies it is also the land planning policy specifically wants developed first, precisely to take pressure off farmland and green belt at the city’s edge. A policy of protecting disused lots for their rare beetles and colonising plants, applied indiscriminately, would not conserve nature so much as redirect the same housing need onto greenfield sites that are usually far worse for biodiversity than the brownfield land being spared. Treating every weedy yard as a reserve is not an ecological argument. It is a way of losing an argument about shelter while feeling virtuous about insects.

The objection is right about the danger and wrong about what follows from the ecology. Nothing in the evidence above supports fencing off every vacant lot against development; the simulation that found the highest conservation value from rotation found it precisely because sites were being redeveloped on a schedule, not preserved (Kattwinkel et al., 2011). What the research actually supports is treating the stock of idle land across a city as a resource to be sequenced rather than either cleared on sight or locked away, keeping some fraction of it in the disturbed, unmanaged condition that produces the biodiversity at any one time while the rest moves through redevelopment. That is compatible with building at the pace a housing shortage demands. It is not compatible with doing what both developers and well-meaning restoration schemes tend to do by default, which is to remove the disturbance from every available site at once, on the reasonable-sounding theory that tidiness and planting are always an improvement on neglect. The lot does not need to be saved. It needs, for a while, to be left exactly as useless as it looks.