Better Together: Evaluating the Complementarity of Earth Embedding Models

2026-05-18

AI Summary

Existing Earth embedding models are often evaluated in isolation, overlooking the potential gains from model fusion. This work presents the first systematic assessment centered on complementarity, evaluating AlphaEarth, Tessera, GeoCLIP, and SatCLIP across six downstream tasks under single-model, pairwise, and joint fusion settings. The study introduces a general Embedding Complementarity Index and demonstrates that fused models significantly outperform the best individual model in four of the six tasks. It further reveals that complementarity is influenced by task type and geographic location, and in land cover regression tasks, it also correlates with the spatial scale of surface features.

Abstract

Earth embedding models transform Earth observation data into embeddings uniquely tied to locations on the Earth's surface. These models are typically evaluated in isolation, comparing the downstream task performance across different Earth embeddings. However, spatially aligned embeddings can naturally be fused, providing richer information per location, a capability that isolated evaluations fail to capture. We therefore propose assessing Earth embeddings by their complementarity: the performance gain of fused embeddings over the best single-model baseline. To operationalise this, we introduce an embedding complementarity index applicable to any embedding and task, and evaluate four Earth embedding models (AlphaEarth, Tessera, GeoCLIP, SatCLIP) in isolation, in all pairs, and jointly across six downstream tasks. Fused embeddings outperform the best single model in four out of six tasks, confirming that single-embedding evaluations often underestimate Earth embedding capabilities. Complementarity proves both task- and location-dependent. Further, for a land cover regression task, we find that complementarity is partially determined by the spatial scale of land cover classes. Complementarity reframes Earth embeddings: the greatest future gains may come not from any single Earth embedding model, but from combinations that are better together.

Problem

• Earth embedding
• complementarity
• embedding fusion
• downstream tasks
• model evaluation

Innovation

• Earth embedding
• complementarity
• embedding fusion
• downstream task evaluation
• spatial scale