AlphaEarth embeddings improve crop classification but limit cross-region use
Evaluating AlphaEarth Embeddings for Crop Classification: Separability, Temporal Dynamics, and Cross-Regional Transferability
Computer Vision and Pattern RecognitionMachine Learning
Summary
Identifying different crops from satellite images is tricky because traditional methods mostly look at raw colors and simple patterns. The authors studied a new way of representing land data called AlphaEarth embeddings, which capture more complex details and crop changes over time, making it easier to distinguish similar crops like corn and soybeans. However, these embeddings don’t work as well when used in a very different region, like transferring from the US to India. Mixing these embeddings with traditional features offers better overall results for monitoring crops.
What this means in practice
- •For agricultural monitoring teams: Use AlphaEarth embeddings combined with traditional features to improve crop type identification and understand crop rotation patterns over multiple years.
- •For regional agricultural extension services: Adopt feature combinations that include harmonic and NDVI-based data for better crop classification when applying models trained in one geographical area to another.
Authors
Priyanka Goyal, Maitreyee Sar, Adway Mitra, Manjira Sinha
Abstract
Earth Observation (EO)-based agricultural monitoring is shifting from pixel-level spectral analysis toward representation-driven modeling, yet systematic evaluation of learned representations across spatial, temporal, and domain-generalization dimensions remains limited. Geospatial foundation models generate high-dimensional embeddings encoding spatio-temporal land-surface properties, but their utility for agricultural applications remains insufficiently understood. This study presents a representation-agnostic evaluation framework integrating geometric separability, temporal dynamics, spatial robustness, and cross-regional generalisation. A 64-dimensional AlphaEarth embedding space is benchmarked against Sentinel-2 reflectance, NDVI-derived phenology, and harmonic features using field-scale data from the Conterminous United States (CONUS; 2017-2023) and Northern India (2021). Traditional features achieve strong linear separability (Fisher ratio), while embeddings provide superior distribution-aware separation -- Bhattacharyya and Mahalanobis distances up to an order of magnitude higher for spectrally similar pairs (e.g. Corn/Soybeans: 3.9 vs. 0.10). Embeddings uniquely capture multi-year crop rotation dynamics, reaching a separability ratio of 1.784 versus a maximum of 1.003 across all engineered features, and exhibit lower residual spatial autocorrelation of errors (mean Moran's $I \approx 0.17$). However, harmonic and NDVI features show stronger cross-continental transferability (US $\to$ India Macro F1: 0.493 vs. 0.326 for embeddings), reflecting better distributional alignment. The combined feature space yields balanced performance, supporting hybrid modeling. These findings reveal a fundamental trade-off between representational richness and cross-region transferability, providing a structured framework for feature selection in large-scale crop classification.