# From Foundation Embeddings to Cropland Maps: Label Efficiency, Temporal Transferability and Independent Human Validation

arXiv: https://arxiv.org/abs/2609.17138 · HTML: https://mlsanity.com/paper/2609.17138/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 30 references that are in the map.
- First posted on arXiv 15 Sep 2026.

## Builds on (most cited first, 9 of 9)

- [XGBoost: A Scalable Tree Boosting System](https://mlsanity.com/paper/1603.02754/) (2016)
- [Foundation Models for Generalist Geospatial Artificial Intelligence](https://mlsanity.com/paper/2310.18660/) (2023)
- [AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data](https://mlsanity.com/paper/2507.22291/) (2025)
- [TerraMind: Large-Scale Generative Multimodality for Earth Observation](https://mlsanity.com/paper/2504.11171/) (2025)
- [AgriFM: A Multi-source Temporal Remote Sensing Foundation Model for Agriculture Mapping](https://mlsanity.com/paper/2505.21357/) (2025)
- [Harvesting AlphaEarth: Benchmarking the Geospatial Foundation Model for Agricultural Downstream Tasks](https://mlsanity.com/paper/2601.00857/) (2025)
- [Cropland Mapping using Geospatial Embeddings](https://mlsanity.com/paper/2511.02923/) (2025)
- [Embedding -based Crop Type Classification in the Groundnut Basin of Senegal](https://mlsanity.com/paper/2601.16900/) (2026)
- [Characterizing AlphaEarth Embedding Geometry for Agentic Environmental Reasoning](https://mlsanity.com/paper/2604.18715/) (2026)

## Abstract (arXiv)

Geospatial foundation models provide reusable representations of satellite imagery that support downstream mapping with limited task-specific modelling. We evaluate whether annual AlphaEarth embeddings support binary cultivated-versus-non-cultivated mapping in Maine, USA, using 192 spatially separated patches and labels derived from the USDA Cropland Data Layer (CDL). Without fine-tuning the foundation model, a lightweight classifier reaches 93.7% overall accuracy and 90.8% balanced accuracy on held-out patches. Logistic regression is within 0.3 percentage points of a gradient-boosted ensemble, while a nearest-class-centroid rule, which uses class centroids but fits no parameters, reaches 90.2%. A balanced sample of 60,000 labelled pixels is within 1.3 percentage points of the full pool of 8.6 million pixels; because pixels are spatially autocorrelated, this result concerns pixel-sample efficiency rather than 60,000 independent annotation sites. In a same-region transfer experiment, classifiers trained in one year remain accurate across 2018 to 2023. Against a blind, two-interpreter consensus at 385 randomly sampled points in one contiguous 2023 block, the AlphaEarth-plus-random-forest map agrees at 95.3% (kappa=0.82), compared with 91.7% for the CDL (kappa=0.72; exact two-sided McNemar p=0.0161). This local result is consistent with partial smoothing of CDL label noise, but it does not establish statewide correction of the reference product. On the same points, the difference from a fine-tuned TerraMind segmentation model is not statistically significant (95.3% versus 93.5%; p=0.14), and the experiment is not a controlled comparison of computational cost. These results support frozen geospatial embeddings as a low-compute candidate for regional cropland mapping, subject to the limits of a single-state study, a 30 m-derived training reference, and a one-block human validation.