The crop after

Part of a crop's value is what it leaves behind for the next one. We almost never count it.

A yield map tells you what one crop produced in one season. It says nothing about the nitrogen a legume left in the soil, the weed seeds a dense cover crop kept from germinating, or the disease cycle a rotation broke. Those effects show up a season later, in a different crop, often on a different budget line.

I learned this from fish, years before I thought much about soybeans or maize.

What the fishponds taught me

In a carp polyculture pond, the waste of one species feeds another, and the whole system is built around that.

My PhD was about integration within the farm: how the byproducts of one activity become the inputs of another. I worked with family farms in Misiones, in northeastern Argentina, where fish ponds were the hub. Pig manure and cassava residues, which would otherwise be waste, went into the ponds as fertilizer and feed.

What made those inputs go so far was the fish. Most farms stocked a mix of carps, a system developed in China over centuries. Each species lives in a different part of the pond and eats something different. Grass carp graze on plants and leave droppings full of undigested fiber. Those droppings fertilize the water, and the plankton that blooms feeds silver and bighead carp near the surface. Common carp root through the bottom, stirring up sediment and putting nutrients back into circulation.

The Chinese literature quantifies these handoffs. An FAO review of Chinese pond culture reports that about 100 kg of grass carp droppings supported roughly 10 kg of plankton-feeding carp downstream (FAO). Each species passes the nutrients along to the next.

So the chain ran from the pigsty and the cassava field, through the pond, to Easter. On most farms the fish were harvested and sold during Holy Week, when demand peaks, and a large share was eaten at home (Sobre La Tierra, FAUBA). The pond sat in the middle of the farm's economy, turning the leftovers of the pigs and the crops into food.

We had a saying in the field. On the farm, the cow eats first, then the pig, then the chickens, and the fish get forgotten, because fish don't make noise when they're hungry. We laughed, but it described a real problem. The pond lived on the farm's leftovers, which made it efficient, and that also made it the first thing to be neglected, because it couldn't complain.

The same logic, on land and in time

A rotation is a polyculture stretched over time. Instead of species sharing a pond at once, crops share a field one after another, and each hands something to the next.

You can measure the handoff. A meta-analysis of field studies found that maize after a legume cover crop yielded 30% to 33% more when nitrogen rates were low or the field moved to no-till; grass cover crops, on average, neither raised nor lowered maize yield (Marcillo and Miguez, 2017). The reverse also holds: crops grown in short rotations or monoculture often yield less than the same crop in a longer rotation, a decline usually traced to pathogens and other soil organisms that build up when the same host returns (Bennett et al., 2012).

When I worked at the California Strawberry Commission, one idea excited me more than most: planting wheat as a cover crop to tip the soil against strawberry diseases. Later trials on the Central Coast tested it against Macrophomina root rot, using wheat cultivars such as 'Summit 515'. The wheat did shift the soil microbiome, raising several bacteria known to attack pathogenic fungi. But in greenhouse trials and the first field season, it did not measurably reduce disease (Steele, 2023; Steele et al., 2023). So the soil changed, but nobody has yet shown that the change pays off on a farm.

Synergy can fail, and Argentina's farmer networks have been honest about it. In a 2018–19 CREA trial in northern Buenos Aires, a hairy vetch cover crop nodulated poorly, drew on soil nitrogen instead of adding it, and the maize after it yielded less than after a bare fallow. Grass cover crops left 20 to 85 mm less water in the profile at soybean sowing and cost about 150 kg/ha of soybean, unless they were grazed (CREA, 2019). A trial in Coronel Suárez found no difference in the following maize at all (Romero, 2023).

A much larger dataset from the US tells a similar story. Using satellite imagery across about 20 million acres of the US Corn Belt, Jillian Deines, David Lobell and colleagues compared fields with at least three years of cover crops against similar fields without them. Maize yielded about 5.5% less and soybean about 3.5% less. The authors traced much of the penalty to competition for nitrogen and water, and expect smaller losses where water is less limiting (Deines et al., 2022).

I don't read this as evidence against cover crops. I read it as evidence against sentences like "cover crops are good." Ken Giller and colleagues have spent years arguing against farming labels that carry a verdict. Conservation agriculture was promoted as a panacea for African smallholders, even though the crop residues it needs as mulch are also the feed for their livestock (Giller et al., 2009). "Regenerative" agriculture is defined so loosely that its practices get pushed "with little regard to context" (Giller et al., 2021). Every practice comes with trade-offs, and how it turns out depends on the soil, the water, the nitrogen and what else the farm needs from that field.

The ponds had already taught me this. Common carp help recycle nutrients from the bottom, but stock too many and they compete with the other species for the same food, which is why Chinese farmers cap their numbers (FAO). The other classic failure was too much of a good thing. Put too much manure in the pond and the water turns green with algae. By day they produce oxygen; at night they consume it, and the fish can run out of air before dawn. A legume that fails to nodulate, or a cover crop that drinks the water the next crop needed, is the field version of an over-fertilized pond. Complementarity has to be designed: which species, in what proportion, and when.

Breeding for the crop after

If complementarity has to be designed, cover crops are the place to start, because their only job is to help the crop after.

Gervasio Piñeiro and colleagues at the University of Buenos Aires helped frame this idea in Argentina. In 2014 they proposed calling these plants cultivos de servicio, service crops, and judging them by the services they provide: holding soil, fixing nitrogen, using water that would otherwise be lost, suppressing weeds. They argued that success depends on capturing energy the cash crops leave unused, and called for research on species and mixtures with better rooting, water use efficiency, nitrogen fixation and nutrient capture (Piñeiro et al., 2014).

Farmer networks in Argentina have since collected a lot of evidence. Across 63 site-seasons in Aapresid's service crop network, a long fallow lost 271 to 409 mm of water to evaporation and deep percolation, water that never reached the next crop. Service crops turned that water into growth instead. Grasses sometimes cut maize yield when the water was not replenished; legumes sometimes raised it; mixes were neutral or positive (El Tribuno, 2026).

Denison's work suggests a next step. Most cover crops are still varieties bred for grain or forage, chosen for how much they grow themselves. But a cover crop's success is measured in the following crop, and natural selection never optimized for that. A grass that drinks the profile dry in a dry spring is doing exactly what selection taught it. Denison points to one target breeders could pursue instead: legumes whose host sanctions reward the most efficient nitrogen-fixing rhizobia, enriching the soil with better partners for every legume that follows (Denison, 2015). The same logic could apply to roots that leave channels for the next crop, residues that release nitrogen when the next crop needs it, and growth that slows before water runs short.

In the ponds, we chose species for what their waste would feed. In a rotation, we could breed cover crops for what they leave behind.

Why we undercount it

We measure crops one season at a time, so the handoff between them falls through the cracks.

Agriculture has a favorite line about this: a farmer gets about 40 chances in a lifetime, 40 growing seasons to get it right. The point about urgency is fair. But the arithmetic hides an assumption, that the 40 seasons are 40 independent tries, like rolls of a die.

They aren't. Each season inherits the weed seeds, pests and nitrogen the last one left, and passes its own leftovers on. Forty linked seasons behave very differently from forty separate tries. A bad decision in year three is still being paid for in year seven, and a good one keeps paying.

A yield map credits the maize, not the vetch that came before it. A field trial usually ends at harvest. A budget books the cover crop as a cost this year and the benefit, if any, as luck next year. Where land is rented season by season, the farmer who pays for the cover crop may not be the one who harvests what it leaves behind.

This is starting to change, and satellites are part of why. Using crop-type maps and satellite yield estimates across major breadbaskets, David Lobell's group at Stanford found that the benefits of a good preceding crop are large but variable, and depend on the weather: bigger in rainy years, and for legume precrops, smaller when it is hot (Kluger et al., 2025). A single field trial can't give that kind of answer. A pond farmer, in a way, already works with it by eye.

Managing the sequence

In my last essay I argued that the population, and not the single plant, is the unit of a crop. This is the same argument stretched over time: the unit is the sequence of seasons.

In a carp pond nobody asks which species was most profitable on its own. The question is what the pond produces, and the answer depends on how the species feed each other. A rotation deserves the same question. I care less about the best soybean or the best maize than about what the field produces across the sequence, and how much each crop leaves for the one that follows.

I spent years away from those ponds, working on conventional systems in California. I keep finding that the most useful idea I brought with me came from them: look for what one part of the system throws away, and design something to eat it.