Do Foundation Model Embeddings Improve Cross-Country Crop Yield Generalisation? A Leave-One-Country-Out Evaluation in Sub-Saharan Africa

TL;DR

This study assesses whether geospatial foundation model embeddings enhance cross-country maize yield predictions in sub-Saharan Africa, finding that they do not outperform traditional spectral features due to yield distribution shifts.

Contribution

It provides a rigorous evaluation of foundation model embeddings versus spectral features for cross-country yield prediction, highlighting the limitations caused by distribution shifts.

Findings

Foundation model embeddings do not outperform spectral features in cross-country prediction.

All feature sets perform poorly under leave-one-country-out validation, with negative R^2.

Main limitation is yield distribution shift between countries, not representation quality.

Abstract

Accurate predictions of smallholder maize yields across national boundaries are critical for food security planning in sub-Saharan Africa, yet most published benchmarks report within-country performance that overstates true generalisability. This paper evaluates whether geospatial foundation model embeddings, specifically Prithvi-EO-1.0-100M and ViT-Base, outperform traditional Sentinel-2 spectral features under a Leave-One-Country-Out cross-validation scheme on 6,404 maize field observations from five African countries. The results show a clear generalisability gap: within-country random cross-validation yields moderate R^2 values, but all feature sets perform poorly under cross-country testing, with universally negative R^2. Frozen Prithvi-EO embeddings provide no meaningful advantage over engineered spectral features for cross-country prediction in this setting. The paper argues that the main limitation is a shift in yield distribution between countries rather than representation quality and releases a reproducible negative benchmark for future work.