From rows to yields: how foundation models for tabular data simplify crop yield prediction.

Accurate and timely crop yield forecasts are crucial for guiding decisions on resource allocation, market strategies, and food security interventions, especially in food-insecure regions. These forecasts enable policymakers and stakeholders in the global food supply chain to anticipate production levels and implement effective mitigation and response measures. As the complexity of data and the need for precise predictions continue to grow, there is an increasing demand for models that can efficiently process and analyze diverse datasets. To address these challenges, we present an application of a foundation model for small- to medium-sized tabular data (TabPFN), to sub-national yield forecasting task in South Africa for maize, soybeans and sunflowers. We used the dekadal (10-day) time series of Earth Observation (EO; FAPAR and soil moisture) and gridded weather data (air temperature, precipitation and radiation) to forecast the yield of summer crops at the sub-national level. The crop yield data were available for 23 years and for up to 8 provinces. Covariate variables for TabPFN (i.e., EO and weather) were extracted by region and aggregated at a monthly scale. We benchmarked the results of the TabPFN against six ML models and two baseline models. Leave-one-year-out cross-validation experiment setting was used in order to ensure the assessment of the model’s capacity to forecast an unseen year. Results showed that TabPFN and ML models exhibit comparable accuracy, both outperforming the baseline approaches. For maize, the best-performing ML model achieved 6.8% rRMSEp, while TabPFN achieved 8.8%, with R² of 0.91 and 0.86 respectively at the national level. Nonetheless, TabPFN demonstrated practical utility due to its faster tuning time and reduced requirement for feature engineering.