A peanut rotation lost 0.53 grams of soil organic carbon per kilogram relative to continuous winter wheat–summer maize across thousands of working farms in Henan Province, China. Rice-inclusive systems gained 0.78 grams per kilogram over the same baseline. The study, published in Plant and Soil, overturns the assumption that diversifying a rotation automatically builds soil carbon.

Xianjin Xie and Anning Zhu of the Institute of Soil Science at the Chinese Academy of Sciences led the work, with colleagues from the Soil and Fertilizer Station of Henan Province and Henan Agricultural University. They analysed soil testing records from 2006 to 2020, drawn from China's Soil Testing and Fertilizer Recommendation program, and paired them with plot-level cropping histories and a modeled soil organic carbon baseline representing the 1980s. That gave them carbon-change data from real farms where farmers, not researchers, made every management decision.

Henan sits at the heart of the North China Plain, where intensive double cropping dominates and groundwater is under strain. The team compared four rotation systems: the classic winter wheat–summer maize rotation (WM), a wheat–maize/soybean system (WML), a wheat–maize/peanut rotation (WMP), and rice-inclusive systems (RIS).

Raw province-wide averages told one story. Every system gained carbon since the 1980s, reflecting rising yields, heavier fertilizer and residue inputs, and policy interventions. The unmatched means were 4.15 grams per kilogram for wheat–maize, 3.73 for wheat–maize/soybean, 3.30 for wheat–maize/peanut, and 3.72 for rice-inclusive systems. On those numbers, the conventional double crop looked best.

The matched analysis told a different one. Farms growing different rotations differ systematically in soil type, climate and management intensity, so the researchers used within-county matching to pair otherwise comparable fields, then fitted paired fixed-effects models to isolate the rotation effect. Once confounders were stripped out, soybean showed no statistically distinguishable difference from pure wheat–maize. Peanut underperformed by 0.53 grams per kilogram. Rice-inclusive systems came out on top at plus 0.78 grams per kilogram.

The divergence between the raw and matched numbers is the study's methodological warning: naive comparisons across an entire province would have produced entirely wrong conclusions about which rotations build carbon.

Soil decides

The most consequential finding concerns soil texture and moisture. The peanut rotation's carbon penalty was sharpest in fields with low clay content and drier conditions, where sandy, well-aerated soils let microbes decompose returned organic matter before it can stabilise. The rice advantage was greatest in high-clay, wetter fields. Clay minerals bind organic molecules and protect them physically, while flooded, oxygen-limited paddy conditions suppress microbial respiration and slow oxidative breakdown. Alternating anaerobic conditions and iron redox chemistry in waterlogged rice soils add further preservation mechanisms documented in paddy research.

This reframes carbon sequestration as a coordination problem between carbon supply and carbon preservation, not an input-output ledger. Legumes, championed in global meta-analyses for their supposed carbon benefits alongside nitrogen fixation, may return substantial residue to a dry, sandy soil only to lose carbon as fast as it arrives. A rice system with modest inputs can still accumulate more carbon because the soil environment locks it away.

The practical read for farmers and advisers is blunt: blanket recommendations to diversify rotations for carbon sequestration will not deliver uniform benefits and may backfire on light, dry soils. Rotation choices should match soil conditions. Legume rotations with peanut stand on their yield and economic case, not carbon claims, on droughty ground. Rice-inclusive rotations on clay-rich, wetter land offer a genuine sequestration opportunity.

The study also demonstrates the power of administrative datasets. By combining soil testing records with propensity-score-style matching and fixed-effects modelling, the team approximated the rigor of a controlled trial at a scale no experimental network could match. The underlying datasets are restricted, but derived data and code are available on reasonable request. The work was funded by China's National Key Research and Development Program.

As Xie and colleagues conclude, crop diversification alone did not guarantee greater soil organic carbon accumulation; the rotation effect depended on the coordination between plant-derived carbon return and soil conditions favoring preservation. For anyone hoping a simple change in crop sequence will bank carbon, the message from thousands of real farms is clear: the soil decides. Watch for follow-up work testing whether these texture-and-moisture interaction effects hold in other major double-cropping regions.