Yield

Many countries report crop yields only for large administrative units, so models trained on coarse records must be applied at finer scales, where few samples and scale-dependent relationships degrade accuracy. A machine learning model trained on field-level soybean yields in the US Midwest was fine-tuned with Brazilian state-by-year records, a technique called transfer learning, then used to predict municipal yields from 2001 to 2021. Transfer across scales rather than across regions had rarely been tested. Without any municipal training data, explained yield variance rose from 0.29 to 0.44, though transfer added little where fine-scale data were plentiful.

Much of what is known about smallholder farming rests on farmers’ own knowledge, and gaps between them and instruments used for measurements are usually treated as error. Reported field size, harvest and season timing from 640 maize farms in Zambia were checked against GPS boundaries, weighed harvests and satellite rainfall. The discrepancies were structured rather than random: small fields were overstated and large ones understated, while low harvests were understated and high ones overstated. Read instead as conceptual mismatch, they point to farmers and researchers holding different mental models.

While potential maize yields in Zambia are 9 t per hectare (t/ha), the average farmer produces only 1–2. We evaluate the contribution of responses to weather risk to that gap by decomposing the yield gap in maize in Zambia. Risk aversion explained 24% of the average gap, varying widely by location, implying 6.75 tons per hectare attainable without added risk against 8.84 when maximizing profit.