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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Adjei, Yaw Osei
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915994885160960
author Adjei, Yaw Osei
author_facet Adjei, Yaw Osei
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2605_08113
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Do Foundation Model Embeddings Improve Cross-Country Crop Yield Generalisation? A Leave-One-Country-Out Evaluation in Sub-Saharan Africa
Adjei, Yaw Osei
Machine Learning
Computer Vision and Pattern Recognition
I.2.6
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.
title Do Foundation Model Embeddings Improve Cross-Country Crop Yield Generalisation? A Leave-One-Country-Out Evaluation in Sub-Saharan Africa
topic Machine Learning
Computer Vision and Pattern Recognition
I.2.6
url https://arxiv.org/abs/2605.08113