Benchmarking Geospatial Foundation Models for Agriculture Applications
2026-06-29 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors tested three big satellite image models (Prithvi, SpectralGPT, and SatMAE) to see how well they work on farming tasks in different U.S. states they weren't trained on. They found that all models struggled to identify less common crops when moved to new regions, often just guessing the most common ones. They also noticed that changing how input data is given to the models affects their performance differently, making comparisons tricky. The authors suggest that testing these models in region-specific ways is important for improving their use in agriculture.
Geospatial foundation modelsSatellite imageryCrop segmentationChange detectionRegional distribution shiftRemote sensingMulti-temporal analysisModel transferabilityAgriculture monitoringBenchmark evaluation
Authors
Zhuocheng Shang, Sanmay Das, Ahmed Eldawy
Abstract
Geospatial foundation models pretrained on satellite imagery promise broad generalization across remote sensing tasks and regions, but their geographic transferability has not been systematically tested, especially in agriculture applications. This paper presents a controlled benchmark that evaluates three models, Prithvi, SpectralGPT, and SatMAE, on multi-temporal crop segmentation and change detection across four U.S. states, Iowa, North Carolina, California, and Minnesota. By assigning each train, validation, and test split to a separate region, we measure how well each model transfers to land it has not seen. All three degrade sharply under regional distribution shift, predicting only the most common crops while missing rare ones. We further find that fitting these models to a shared input format affects each one differently, which complicates direct architectural comparison. These results expose key limitations of current geospatial foundation models for agriculture and point to region aware evaluation as a necessary standard.