The land saved by mixing maize and soybeans depends on where they grow

A field can favour a partnership that a whole continent makes less compelling. Maize and soybeans grown together can use land efficiently, but growing them separately gives a region more freedom over where to put each crop. The advantage of sharing a field has to survive that geographical comparison.[1][2]

Mathilde Chen, Nicolas Guilpart and David Makowski examine that question in a Nature Communications study published on 6 October. They combine machine-learning estimates of European crop yields with evidence from intercropping trials, then calculate how much land different production strategies require. The result is a conditional land-use comparison, built around specified harvests and farming assumptions.[1]

Two harvests from one area

The starting agronomic assumption is unusually easy to picture. In the study’s baseline, one hectare of maize–soybean intercropping produces as much soybean as 0.56 hectares of soybean grown alone, and as much maize as 0.79 hectares of maize grown alone. The two separate areas add to 1.35 hectares. These are yield equivalents, not the proportions of a field assigned to each plant.[1]

That arithmetic implies about 26% less land for the mixed system when the comparison stays within the same growing conditions: one hectare replaces 1.35. But the ratio cannot settle a European allocation problem by itself. If a location is particularly good for soybeans and relatively poor for maize, growing both there commits the maize to that disadvantage. Separate production can change the balance.[1][2]

This is where the machine learning enters. Separate random-forest models learned relationships between climate, water management and historical yields, using 116,516 soybean site-years and 72,964 maize site-years from major producing regions worldwide. A site-year is one location in one year, so repeated years at a location are part of those totals. The researchers adjusted historical yields to a 2016 technology baseline and estimated rainfed yields for 2,699 European grid cells using the weather of 2000–2023.[1]

The models supply the geographical variation in yields. They do not learn how the two species interact in a mixed field. That second ingredient comes from the field-trial yield ratios, applied to the predicted sole-crop yields. In particular, the baseline does not let those ratios change with local weather; the paper identifies insufficient intercropping data as the reason. European soybean observations were also sparse in the yield-model training data.[1]

Keep the maize in the comparison

Consider the scenario targeting 18.2 million tonnes of soybean, half the study’s 2018–2022 soybean-equivalent demand baseline. The allocation procedure prioritises grid cells with high potential soybean production. It allows the crops to return once every four years, represented by using up to a quarter of a cell’s cropland each year. This is a soybean-prioritised allocation, rather than a claim to have found the globally optimal arrangement of every crop.[1][2]

Intercropping reaches that soybean target on 13.1 million hectares while also producing 63.7 million tonnes of maize. The separate-crop comparison first allocates enough land to soybean in the most productive cells, then assigns maize to the residual area. Matching the soybean alone needs 7.1 million hectares. Keeping total area at 13.1 million hectares leaves six million for maize, whose harvest is then much smaller.[1][2]

Restoring the missing maize requires another 3.4 million hectares, taking the separate-crop total to 16.5 million. The saving is therefore 3.4 divided by 16.5, about 21%. Comparing only the soybean harvest would have concealed the advantage that comes from producing the companion maize.[1]

A saving against which alternative?

The central 13.1-million-hectare intercropping scenario is still larger than the roughly ten million hectares used for maize and soybean in the paper’s 2018–2022 reference period. Its 21% saving is relative to a separate-crop system producing the same larger pair of harvests. It does not mean the proposed expansion would release that fraction of today’s cropland.[1]

Timing also matters. When the researchers assumed both crops were sown and harvested together, the field-derived yield ratios fell to 0.53 for soybean and 0.76 for maize. Repeating the matched-production comparison at the 50% soybean target reduced the estimated land saving to about 18%. The advantage persisted, but the way the partnership was managed changed its size.[1][2]

Harvesting the partnership creates another practical task: separating its products. The paper identifies specialised sowing, harvesting and sorting machinery, along with grain-purity requirements, among the obstacles to wider adoption. A favourable land calculation does not supply those operations. It does show what their successful organisation could preserve: two substantial harvests from an area smaller than the corresponding separate-crop alternative.[1]

The geographical calculation changes the field-scale answer from roughly 26% to roughly 21% in the central scenario. What the yield models add is a way to count the cost of placing the crops together across an uneven landscape. Under the study’s assumptions, sharing fields still comes out ahead after that cost is counted.[1][2]

Report an error or suggest a correction