The Slow Violence of Soil

Sometime around 2100 BCE, the Akkadian Empire stopped receiving grain shipments from its northern provinces. This was not a policy failure or a logistical breakdown. The grain had simply stopped growing. Sediment cores pulled from the beds of ancient Syrian lakes in the 1990s by geologist Harvey Weiss and his colleagues show a sudden shift: a two-to-three-century drought, almost certainly exacerbated by the deforestation and intensive agriculture that the empire had used to feed its expansion. The Akkadians had eaten their way through the agricultural surplus that built them, then through the soil that generated the surplus, then through the climate stability that the soil maintained. When the rains finally returned, the empire did not.

This sequence is not a curiosity. It is a pattern that repeats across four thousand years of recorded history with a regularity that should be alarming and is instead mostly ignored.

The reason it gets ignored is largely a matter of timescale. Soil degradation operates on a schedule that is invisible to individual human lives and mostly invisible to the political cycles that govern collective decisions. A field that is being destroyed by erosion looks, in any given growing season, very much like a field that is being farmed sustainably. The yield might be slightly lower than it was a decade ago, or the inputs required to maintain that yield might be slightly higher, but neither signal is dramatic enough to force a reckoning. The reckoning comes later, sometimes much later, and by then the people who made the original decisions are long gone.

Jared Diamond, whose 2005 book Collapse mapped this pattern across multiple civilizations, described it as a process of “creeping normalcy” — the way that each generation calibrates its sense of normal to the degraded baseline it inherits, so that nobody ever perceives the full magnitude of the change. The Easter Islanders who cut the last tree presumably did not think of themselves as cutting the last tree. They thought of themselves as cutting a tree, because that was what their parents had done, and their grandparents, and by the time the population of trees had declined to the point where the last one was imminent, the population of people who remembered a different forest was also gone.

This is not stupidity. It is the ordinary operation of human cognition applied to a problem that human cognition evolved very poorly to handle.

The Dust Bowl of the 1930s offers a more recent and more thoroughly documented version of the same failure. The Great Plains had been broken by the plow in a 30-year agricultural rush that followed the Homestead Act and the arrival of the railroads. The native grasses that held the soil had been replaced by wheat, which was enormously profitable during the wet years of the early twentieth century and catastrophically fragile when the rains stopped. When the drought came in 1931, there was nothing to hold the plains in place.

The clouds of dust that darkened Kansas City in April 1935, that deposited grit on the desk of Franklin Roosevelt in Washington, that sent 3.5 million people walking west toward California, were the accumulated consequence of decisions that each seemed reasonable in isolation: plow the grassland, plant the wheat, sell the wheat, plow more grassland, plant more wheat. Nobody decided to cause a catastrophe. They decided, repeatedly, to take advantage of an opportunity. The catastrophe was the arithmetic consequence of the decisions taken together.

What is striking about the Dust Bowl, from the vantage of 2029, is how thoroughly it was responded to. The New Deal’s Soil Conservation Service, established in 1935, was a serious federal investment in teaching farmers to think about soil as a capital asset rather than a free input. Windbreaks were planted. Contour plowing was promoted. The Soil Conservation Districts gave farmers institutional structures within which to coordinate practices that none of them could implement alone. By the 1950s, the Great Plains were largely stabilized, and the mechanisms that had caused the catastrophe were at least partially understood and partially controlled.

Then, slowly, over the following decades, the institutional memory faded, the political support for conservation thinned, and the market incentives — which had never aligned with long-term soil health — reasserted themselves.

The USDA has been tracking topsoil loss on American cropland since the 1970s. The numbers are not secret. In 2022, a study in the Proceedings of the National Academy of Sciences by David Pimentel and colleagues estimated that American farms were losing topsoil at roughly ten times the rate of natural replenishment. Globally, the Food and Agriculture Organization of the United Nations has repeatedly warned that at current degradation rates, the world has roughly sixty harvests of viable topsoil remaining in the most intensively farmed regions.

Sixty harvests. That is sixty years, roughly speaking. It is within the lifetime of children alive today. It is a crisis on a timeline that is entirely comprehensible to human cognition, unlike the four-thousand-year sweep of civilizational collapse. And yet it generates considerably less political urgency than the price of soybeans in any given quarter.

The reason, I think, is that the mechanism of soil loss is still invisible even when the numbers are visible. Topsoil does not disappear in a way that is perceptible to someone standing in a field. It thins imperceptibly, year by year, replaced by inputs (fertilizers, herbicides, irrigation) that maintain yields even as the underlying biological complexity of the soil collapses. The field looks fine. The yield is acceptable. The profit margin is squeezed but manageable. The fact that you are farming a progressively more artificial medium, one that will require progressively more intervention to produce the same output, is not a signal that the market price of this year’s crop reliably transmits.

What is actually being lost is not primarily bulk soil material, though that is being lost too. What is being lost is the biological complexity that makes soil productive. A healthy agricultural soil contains something on the order of a billion bacteria per teaspoon, along with fungi, protozoa, nematodes, and a cascading food web of organisms that together perform functions that no synthetic input can fully replicate: cycling nutrients, suppressing pathogens, building the structure that allows water to penetrate rather than run off, and creating the organic matter that gives soil its characteristic dark color and its capacity to retain moisture.

Industrial agriculture treats this complexity as background noise. The operating assumption is that you can strip the biology out of soil, replace its functions with manufactured inputs, and continue to generate acceptable yields. This works, up to a point. It works better than proponents of regenerative agriculture sometimes admit. But it works by burning through a form of biological capital that accumulated over thousands of years and cannot be quickly rebuilt once it is gone. The Mesopotamian soils that supported the Akkadian Empire have never recovered their agricultural productivity. They are desert now, or close enough to make no practical difference.

The historical record on this point is clear. Civilizations that destroyed their soils did not recover. The valleys of the Fertile Crescent, which gave the world its first cities, its first writing, its first legal codes, produced roughly 40 percent of the world’s grain supply in 5000 BCE. They are net food importers now. This is not a coincidence or a quirk of geography. It is the arithmetic of extraction finally coming due.

What makes the current situation different from all previous versions of this story is scale. Every previous civilizational soil collapse was local. The Romans stripped the forests of North Africa and reduced its agricultural capacity; Carthage never recovered; but the farming continued elsewhere, and Rome eventually found other sources of grain. The Mesopotamian collapse was devastating for the Akkadians, but the rest of the world continued to farm. Today, the degradation is roughly synchronous and roughly global. There is no elsewhere to shift to.

There are things that can be done. Cover cropping, reduced tillage, restored perennial vegetation in the right places, managed grazing, rebuilt wetlands and riparian zones — the toolkit of regenerative agriculture is real and tested and capable of rebuilding soil biology on a timescale of decades rather than millennia. The obstacle is not knowledge. The obstacle is incentive structure: soybeans at $12.40 a bushel generate one set of decisions, and environmental outcomes thirty years downstream generate a different set, and the market does not reconcile them.

The civilizations that collapsed through soil loss were not stupid and they were not ignorant of what they were doing, at least not entirely. They were caught in the same trap we are: individually rational decisions producing collectively catastrophic outcomes, the cost always falling on people who had no vote in the original decision.

The slow violence of soil does not announce itself. It just keeps going.

There is a particular kind of grief that attaches to this subject, which is the grief of knowing that we know. We have the sediment cores from ancient Syrian lakes. We have the USDA data. We have the sixty-harvests estimate and the measurements of topsoil depth and the peer-reviewed literature on soil biology. We know what we are doing and roughly what will happen and approximately when the consequences will become impossible to ignore.

The Akkadians did not have this. Their empire collapsed into a failure they could not have named or measured. We will collapse, if we do, into a failure we measured precisely, analyzed thoroughly, published in journals, presented at conferences, and then continued to cause because the quarterly earnings report needed to look a certain way.

This is not a particularly hopeful way to end an essay. But the soil does not offer hope or despair; it offers consequences. The only question is whether we encounter those consequences in time to do something about them, or whether we discover, as so many civilizations have before us, that “in time” was a category that applied to other people in other eras, not to us, not here, not now.

The plain fact is that it applies to everyone, eventually.