The impossible dream of overyielding polyculture

Every field, every movement, has an idea that is simply too good to ignore. The theory is sound, the small-scale experiments work, and everything points toward an impending revolution. And yet when you try it in practice, it just doesn’t scale up, doesn’t pay off, doesn’t work.

For sustainable agriculture, that idea is “overyielding polyculture”.

What is “overyielding polyculture”?

Let’s start with polyculture. You may have heard of “monoculture”: you know, the thing where you grow nothing but corn in a giant field. Polyculture is “not that”: you grow more than one thing together in a single field, a process known as “intercropping”.

The “overyielding” adjective describes the notion that, while you may get less of any one crop from the field, by growing them together you’ll get more in total than if you split the land and grew them side-by-side.

This is not a new idea: you might have heard of the Three Sisters, a famous Native American agricultural practice that used these principles. Maize grows tall, beans grow up it and fix nitrogen, the squash shades the soil, keeping it cool and moist and free of weeds.

Let’s make up some numbers and put them in a table¹ to demonstrate the idea. We have 3 units² of area, and, as savvy farmers, want to maximize the yield. How should we plant our crops?

Experimental Setup	Area, Corn	Yield, Corn	Yield/Area, Corn	Area, Beans	Yield, Beans	Yield/Area, Beans	Area, Squash	Yield, Squash	Yield/Area, Squash	Total Yield
Pure Corn	3	6	2	0	0	N/A	0	0	N/A	6
Pure Beans	0	0	N/A	3	6	2	0	0	N/A	6
Pure Squash	0	0	N/A	0	0	N/A	3	6	2	6
Side-by-Side	1	2	2	1	2	2	1	2	2	6
Intercropped	3	3	1	3	3	1	3	3	1	9

There we are: one beautifully simple table collapsing messy biological reality so you can understand it.

The intercropped setup counts the area as being planted in maize AND beans AND squash: squeezing them all into the same area at a lower planting density. The yield of each crop goes down, yield per area planted in each crop goes down, total yield goes up! The last two lines are the most important: why does planting 1 field with 3 crops tend to produce more than 3 fields with 1 crop each?

Why does intercropping improve yield?

The theory behind overyielding polyculture is both simple and robust. By growing multiple things (plants or animals) you get:

more efficient use of resources (sun, light, space, nutrients) via niche differentiation
temporal variation in when those resources are needed
increased resistance to pests and diseases
natural hedging against poor weather: each crop is resilient to slightly different conditions
economic diversification, in case market prices change

In short: a harmonious ecosystem working together in a beautiful and delicious virtuous cycle. Genuinely, a lovely notion.

More importantly though, the evidence suggests that it works. Like, not “kind of works if you squint at the data in some studies under some conditions”. Like: genuinely produces large yield improvements, reduces correlated risks and is, in my semi-professional opinion, an agronomic best practice on par with “applying fertilizer”, “watering your crops”, “plant breeding” and “crop rotation”.

The standard way to quantify this is Land Equivalent Ratio (LER): the relative land area required under monocropping to achieve the same yield (across all crops, accounted in a single currency) as a polyculture setup. For our extremely fake data above, that works out to 1.5. If you had 1.5 times as much area in each of maize, beans and squash grown independently, your total yield would be the same as that of the intercropped design.

A simple enough metric³, but a genuinely useful one. The literature⁴, taken skeptically, shows that the measured values here are generally somewhere between 1.2 and 1.8, varying heavily based on the exact details.

It is hard to overstate how crazy that effect size is for agricultural yields. Biochar has solid evidence, and increases yields by 14% on average with wide variation, per Han et al. 2023. No-till agriculture is one of the most successful and popular sustainable agriculture movements of the last century. And yet Pittelkow et al. 2015 estimates that on the whole, it reduces yield by about 5%, gaining reduced input costs and better soil health in exchange.

Jiang et al. 2025 estimates that the central idea that powered the entire Green Revolution, of “maybe we should add synthetic nutrients”, gives a 50% yield gain in corn, the most famously nutrient-hungry crop. Tudi et al. 2021 ballparks the global productivity gain of pest control at about 50%.

This is far from a rigorous review of any of these effects, but the point stands: “agronomically revolutionary” is not hyperbole hyping up some crank technique. You get less of your most profitable or caloric crops, but reduce deadly variation. Those kinds of gains are genuinely unmatched for a technique that is not used at scale and can, in theory⁵, be stacked on top of all the other improvements.

Polyculture at home

Polyculture is so effective that it is the standard approach to home gardening, in part because home gardeners are optimizing for different things. Ever met someone who’s growing an entire backyard of sweet potatoes⁶ because that’s the single most efficient way to produce calories in a fixed unit of area? I didn’t think so!

Carefully adjusting your crops to meet the heterogenous space you have available⁷, spreading your harvests across the season, hedging your bets in case one of your crops gets crushed by a cabbage hornworm, taking advantage of nitrogen-fixing crops like beans: these practices are well-established lore among grandmothers worldwide, not “fancy science thing”.

The idea that gardeners are optimizing for a complex basket of ease, pleasure, beauty, yield, input costs and so on has been well-studied in environmental economics. Those secondary benefits are hard to measure, and not captured by the financial value of the crops, despite making the world a better place. Given the freedom to optimize for things beyond income, humans overwhelmingly choose polyculture. Diversity is our strength, etc, except instead of maximizing corporate profit with the power of Woke we’re making salads with more than one ingredient.⁸

The edible forest garden permaculture sickos⁹ take this one level further, obsessing over “guilds” and “ecological goods and services” and “perennials”. But the core idea is everywhere in home gardening once you know to look for it.

How does it fail?

Such a simple idea, with such strong evidence¹⁰ of success! Why then, as you drive through the American Midwest (where, historically, the Three Sisters were planted), do you see giant fields of monoculture corn beside giant fields of monoculture soy? No, it is not because the farmers who own those fields are ignorant of science, bad at finance, and resistant to change¹¹.

The fundamental problem with growing food as a business is that you have to make money.¹² And making money in agriculture is only loosely connected to the whole business of growing food. Unlike home gardeners, farmers do not have the luxury of optimizing for joy.

If the agronomic evidence is strong (and while it’s imperfect, I think it genuinely is), something else must be going seriously wrong.

As a farmer, you need to:

buy seeds (or other propagules)
plant your seeds
apply fertilizers
weed and tend your crops
kill any pests that might threaten your crops
harvest the crop
sell the crops
have insurance in case your crop fails so you don’t immediately go bankrupt from the loans you took out
collect those sweet, sweet government subsidies

Minimizing costs and maximizing profits at every step of the process. Increasing yields is great, but it needs to be carefully weighed against making other parts of the process harder. Rural farmland is shockingly cheap frankly, which makes “do something stupid but scalable” a dominant strategy. The bitter lesson of agriculture, as it were.

The LER numbers that were so impressive don’t matter in practice, because they’re optimizing for the wrong thing. Swapping to polyculture makes every single one of those steps so, so much harder at scale. Just to grow this at scale you need to source more seeds, plant them in a complex strategy at different times of the year, carefully tailor fertilizer and pesticides to the entire collection of plants, watch out for dozens of different pests, being careful not to poison other crops just before you sell them.

Harvesting is an entire separate disaster. You need unpredictable labor forces that require more skill and can’t work as quickly. You need complex storage and transport infrastructure. And you need a whole web of purchasers: one for each of your crops. “Just throw it in the Shipping Bin” is load-bearing for Stardew Valley’s frivolous, joyful polyculture. Turning “tasty, nutritious food” into “cold hard cash” is extraordinarily hard, especially if you want to get good rates for it. Suppliers want predictability and reliability, the simplicity of a single farm that sells a single thing. In short: they want legibility¹³.

The legibility problem gets so much worse when dealing with the government and its vital agricultural subsidies. These subsidies are often carefully targeted to win the votes of specific farmers in specific regions. To get them, you need to be cleanly classified into one of the targeted groups. “How many acres of corn do you have planted” is a question that fits cleanly for Big Ag while the farmers trying to grow food forests slip through the gaps. This problem compounds for crop insurance, another critical form of farm subsidy, which pays you money even if your crops fail. Unusual crops and polyculture systems are fundamentally idiosyncratic, making the actuarial math of insurance much harder, even when the regulatory regimes are not actively hostile to it. If your Scientifically Superior planting strategy is illegible, you can’t claim these subsidies, you can’t compete, and you go broke, even if your yields are genuinely better.

Industrial agricultural equipment is a modern miracle, and it simply cannot cope with the complexity of polyculture in the same way that little fleshy robots in backyards across the nation can. Without mechanization, you’re stuck competing with farmers from the 1800s, and even the very best 1800s farmer simply gets run over by the homogenizing power of the mechanical thresher. In very specific situations: with high land prices and a market willing to pay a premium, you can make it work. But those are the exceptions: permaculture market gardens that sell to farmers markets cannot and will not revolutionize society.

This all makes me very annoyed, because I really like polyculture, and growing lots of delicious food, and the ecological and social and aesthetic benefits that go beyond simple dollar values. Society would be better if we were able to replace industrial monoculture agriculture with permaculture food forests at scale. But wishful thinking does not a profitable business make, and without profitable businesses, you will never displace the endless fields of corn and soy.

But then why do agrovoltaics work?

However, the recent history of agrovoltaics creates a gap in this oh-so-clean tragedy. Agrovoltaics is the ever-so-pretty name for the idea that you can grow food and harvest solar power on the same land.

At first blush, this makes no sense: don’t plants need light to grow? Yes, but: this is the same strategy as polyculture as a whole. Diversify income, add structure, grow vertically and take advantage of the shady corners.

If you’re not a farmer (or a tropical gardener), you may not realize that it’s possible to have too much sun. Especially if you’re growing C3 crops, you’ll get most of the yield with half of the light, and the shade cuts water use while reducing heat stress. Farmers even deliberately use “shade cloth” to selectively shade sensitive crops or survive the peak of summer.

And just like the rest of polyculture: the math works. Pandey et. al 2025 suggests that typical LERs are about 1.5, with side benefits like reduced water usage and diversification of revenue streams. Grazing is a particularly good fit, as the very weeds that need to be controlled are turned into feed while the partially-covered fields create diverse habitat for both plants and wildlife.

However, unlike the rest of polyculture, agrovoltaics is genuinely starting to scale up¹⁴. While slow compared to the adoption of solar power in general, due to regulatory uncertainty, confusion and fragmentation around land use and high capital costs, it is fast and accelerating compared to other forms of polyculture.

And when you dig into the forms of agrovoltaics that are actually catching on (by scale, not feel-good coverage), there’s a revealing pattern: the rest of permaculture was left behind. Instead of Californians planting elegant shade-tolerant guilds under solar panels, ex-diesel mechanics are grazing sheep under solar panels in Appalachia.

Instead of being promoted in zines and bougie permaculture design courses, Shell (yes, the oil company) is touting large scale trials of agrovoltaics for the industrial scale conventional production of wheat, corn and soybeans.

Because instead of finding a new buyer for a new crop, you simply sell the electricity back to the grid, and keep growing the same monoculture you always have. Cash crops come and go, but nothing could be more practical or legible than a fully-automated commodity.

Agrovoltaics has escaped the hippies and is starting to catch on among the normies. Its adoption is a-la-carte: solar panels and polyculture juxtaposed with chemical-drenched mechanized monoculture, double-dipping on government incentives. The critical difference is unsurprising: agrovoltaics are simple, and that simplicity renders it legible. At least, relatively.

Is this the solarpunk future I was promised?

The necessity of practicality

If agrovoltaics teaches us that polyculture fundamentally works, but we need simplicity to make it practical at scale, we should see evidence of effectiveness for other forms of simple polyculture.

The problem here is not just legibility: in many cases, it’s practicality. Not every crop suffers from this: fruit and vegetable production is largely manual, and grazing animals are even smarter than Roombas. They can work around trees and navigate the complex structure of polyculture. But field crops, the cereals and legumes that make up the vast majority of human calories, animal feed, and global agricultural land, flatly require mechanization to be competitive. Any solution that has a chance of scaling needs to accommodate these very physical limitations.

And some of them do. Struggling to produce enough food on limited arable land, China has reimagined the familiar soy-corn complex of the Midwest, pursuing an initiative known as Maize Soybean Strip Intercropping (MSSI). The design is compatible with mechanized harvest and has been widely but not universally adopted: exactly the sort of “emergent, practical technology” story we would hope to see.

Li et al. 2023 report an incredible LER of 2.27 across a multiple-year field study. Despite the cynicism, LER is genuinely meaningful, as long as everything else works out. While land may be cheap, it is neither infinite nor free, and pricy, scarce inputs like water and fertilizers tend to scale linearly with area. Everything else working out is, however, the hard bit.

Can we scale it further? The biology is, to a first approximation, completely the same the world over. The reason this works in Shandong and not Iowa is that imposed legibility has wrought practicality. The Chinese state, in all its great and terrible power, has declared this particular form of polyculture legitimate, warping both the regulatory and economic environment to accommodate it.

The semi-arid Sahel has grown a tremendously important success story of its own, using trees to shade crops under the name of Farmer Managed Natural Regeneration. The Sahel has faced creeping desertification, and unlike China, lacks the economic resources and strong central government required to confront the problem head-on. Instead, farmers have taken the lead, carefully nurturing the natural sprouts of native trees, creating an overstory that stabilizes the soil, lowers temperatures and improves drought resilience. Faidherbia albida is particularly synergistic: fixing nitrogen then dropping its leaves just before the rainy season that the understory crops use to grow.

This directly increased the yields of the understory crops (sorghum, millet and livestock) by something like 20%, while farmers consume many of the side benefits of the trees (medicine, fuel, lumber) themselves, or liquidate them as needed in hard times. No elaborate guilds: just ordinary crops, sheltered by trees. As of 2019, trees cover a staggering 16% of the cropland in the region. Simple, cheap, flexible; and thus effective.

Despite these successes, carbon credits tell a cautionary tale. You can not simply manufacture legibility; it cannot always be imposed. Trees and forests are fantastic self-contained carbon-capture-and-storage factories. Agroforestry advocates were excited: it promises much better carbon capture than conventional agriculture, and so by pricing in that externality we can align incentives and shift the equilibrium point!

Unfortunately, carbon is messy, and not just in powdered form. What does “displacement” mean? What are the time scales the carbon is stored for? How can we actually measure soil carbon? Who’s auditing all this?

All of that mess scared off buyers¹⁵, and prices remain low and volatile. For all its elegance, carbon pricing ultimately did not move the needle for tropical agroforestry. Carbon credits might work, one day, but the market that succeeds will be legible: standardized, verified, commoditized.

Polyculture in plain sight

But the most compelling success story is hiding in plain sight: something that we’ve been doing for millennia. Crop rotation is temporal polyculture, and gives us some fraction of the complementary niche benefits with a corresponding fraction of the complexity. This is part of why the LER numbers we discussed at the beginning of this over-long essay can be misleading: crop rotation is polyculture, and is often the commercial baseline to beat, even if it’s not the one we’re measuring against. Alternating soy and corn gives us free nitrogen fixation, growing multiple crops in the same year uses the otherwise barren soil and wasted light, winter cover crops build organic matter at a fraction of the cost of amendment and rotating what’s planted in your conventional market garden keeps pests and disease under control. Imperfect and inefficient but simple.

Farmers already get to keep mechanization and uniformity while dipping their toes into diversity. Polyculture has not failed to penetrate commercial agriculture; in a very important way it has already won.

And yet: it’s not enough. We can do better: crop rotation was revolutionary, but something is genuinely lost without rich spatial intercropping and long-lived perennials.

Scaling up

Overwhelmingly, the people who make or claim to make money by doing permaculture for crop production aren’t building towards a model that can scale; they are not immanentizing their grand dreams of a sustainable, socialist, solarpunk future.

At their best, they’re honest, profitable small businesses that have found an ecological niche of their own, living the dream off a sympathetic urban elite. Or hobby farms carefully burning inheritances and retirements, teaching fellow travelers, and making the world better through idiosyncratic charity¹⁶.

At their worst, they’re suckering idealistic backpackers into back-breaking work at below the minimum wage, making TikToks about being tradwives or running pyramid schemes teaching people how they too can get rich with permaculture practices.

In the end though, their exact motivations and methods don’t really matter. By pretending that their “success” means that there’s something there, that we’re just one more commune away from breaking through, we can ignore the cruel hard reality of living in a society and the tyranny of scale.

To solve these problems, to actually spread this technology, to change the world, we need to take all of the unpleasant, boring problems seriously: the economics, the logistics, the legibility.

That’s the repeated pattern running through this: farmers seek to understand their crops, purchasers seek to understand their suppliers, governments seek to understand their citizens. And at every step, the rich texture of reality is crushed into tidy little boxes and accounted on spreadsheets. Even as someone who adores spreadsheets (after all, what is a spreadsheet but a table you can play with), I know that they’re fundamentally oppressive lies.

But unfortunately, if we wish to make a difference, we cannot simply opt-out; growing in messy, joyful chaos without a plan, taking every day one breath at a time. Mechanization is flat-out too effective to ignore, and no, you don’t get to claim the high ground standing on the backs of the immigrants picking your strawberries. Operability is legibility; reflections of the same brutal practicality seen through different lenses. And both are a necessary evil of scale, no matter your ideology or ethos.

I hold onto a hopeful synthesis though: we must push the boundaries of what it means to be legible¹⁷. Carefully grow expertise and taste and flexibility in ourselves, then nurture it in all of these institutions, from farm to federation. Statutory reform, precision agriculture, humanoid robotics, workers who learn and grow year-after-year: these are all the same shape of solution at different scales. Rebuild grandma in steel, and steal her wisdom for the good of the world. Dr. Robotnik would be proud.

Start with agrovoltaics, simple intercropping, maybe a bit of agroforestry. Keep experimenting in your backyards and research stations and communes. See what works, share what you learn, and enjoy the journey while looking the destination clearly in the eye.

We must attack the problem on both fronts: reducing complexity without sacrificing the benefits while growing our capacity to understand and manage that complexity¹⁸. With a little luck, a spot of cooperation and a tremendous amount of hard work, we can escape the decades-long dead-end of artisanal agriculture. Imperfect self-similarity is more than an aphorism: quantity has a quality all of its own.

We love tables around here. ↩
The customary units of both crop area and yield are such travesties that I will not dignify them even in a gag! This is a metric household damn it. You will measure your corn harvests in grams and like it. ↩
The definition is simple, but the devil really is in the details. Swapping between calories, kilograms, dollars and “inclusive value based on ecosystem goods and services” completely change the math for both better and worse. The literature is not consistent, because people’s values are not consistent! And keep reading to understand the impact of crop rotation on these comparisons. ↩
Yu et. al 2015 cites a fairly tight average of 1.2 for annual intercropping, Pent 2020 suggests 1.5 for temperate silvopastures, and Chapagain et al. 2018 is a real if dubious study showing an eye-watering 2.45 in a complex tropical system. ↩
As we get into below, the gap between theory and practice is the whole problem here. ↩
The tomato people have something completely different wrong with them. ↩
That’s precision agriculture baby: a related but distinct concept. ↩
Which, for the record, is also Woke. ↩
Hi!! Have I told you about the wonders of Edible Forest Gardens by Dave Jacke and Eric Toensmeier yet? ↩
Yes, yes, I know the comparisons were often apples-to-oranges, the lit review was incomplete, and reality is much more complex and messy. Which are exactly the same mistakes smart, sincere advocates make. Welcome to the essay. ↩
In this case. You’d be surprised how often that is the explanation! ↩
No, we should not return to distributed subsistence agriculture. The Great Leap Forward was bad, and backyard gardens are not any better than backyard furnaces. ↩
The canonical reference on this is Seeing Like A State by James C. Scott, which explicitly examines the polyculture question, but draws the opposite conclusions! ↩
Martínez-Hernández et. al 2026 is a lovely, balanced review that looks at both the technical and regulatory sides of the problem. ↩
In 2023, The Guardian reported that 90% of rainforest carbon offsets by biggest certifier are worthless, quickly tanking and contracting the market. ↩
What a lovely way to retire. Maybe I’ll do that one day, for my own selfish pleasure. ↩
For an academic take on a closely related but distinct problem, read Magrini et al. 2016: “Why are grain-legumes rarely present in cropping systems despite their environmental and nutritional benefits? Analyzing lock-in in the French agrifood system.” Legibility limitations are not the same as legislative lock-in, because the former accepts that irreducible complexity has an irreducible cost. ↩
While the state must see, perhaps we can build it spectacles? ↩