Better Together: Evaluating the Complementarity of Earth Embedding Models

• URL: http://arxiv.org/abs/2605.18667v1
• Date: Mon, 18 May 2026 17:10:13 GMT
• Title: Better Together: Evaluating the Complementarity of Earth Embedding Models
• Abstract summary: Earth embedding models transform Earth observation data into embeddings uniquely tied to locations on the Earth's surface.<n>These models are typically evaluated in isolation, comparing the downstream task performance across different Earth embeddings.<n>We propose assessing Earth embeddings by their complementarity: the performance gain of fused embeddings over the best single-model baseline.
• Score: 5.793071003655904
• License: http://creativecommons.org/licenses/by-sa/4.0/
• 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.

Related papers

• Earth-o1: A Grid-free Observation-native Atmospheric World Model [81.55899748753434]
  We present Earth-o1, an observation-native atmospheric world model that overcomes structural limitations.<n>By integrating diverse sensor inputs into a unified, grid-free dynamical field, the model autonomously advances the atmospheric state in space and time.<n>We show that this fundamentally distinct paradigm enables direct, real-time forecasting and cross-sensor inference without the overhead of explicit numerical solvers.
  arXiv  Detail & Related papers  (2026-05-07T14:27:48Z)
• PRUE: A Practical Recipe for Field Boundary Segmentation at Scale [50.194423500109025]
  We conduct the first systematic evaluation of segmentation and geospatial foundation models (GFMs) for global field boundary delineation.<n>A U-Net semantic segmentation model outperforms instance-based and GFM alternatives on a suite of performance and deployment metrics.<n>Our approach provides a practical framework for reliable, scalable, and reproducible field boundary delineation across model design, training, and inference.
  arXiv  Detail & Related papers  (2026-03-28T02:47:46Z)
• Towards Scalable and Generalizable Earth Observation Data Mining via Foundation Model Composition [0.0]
  We investigate whether foundation models pretrained on remote sensing and general vision datasets can be effectively combined to improve performance.<n>The results show that feature-level ensembling of smaller pretrained models can match or exceed the performance of much larger models.<n>The study highlights the potential of applying knowledge distillation to transfer the strengths of ensembles into more compact models.
  arXiv  Detail & Related papers  (2025-06-25T07:02:42Z)
• TerraFM: A Scalable Foundation Model for Unified Multisensor Earth Observation [65.74990259650984]
  We introduce TerraFM, a scalable self-supervised learning model that leverages globally distributed Sentinel-1 and Sentinel-2 imagery.<n>Our training strategy integrates local-global contrastive learning and introduces a dual-centering mechanism.<n>TerraFM achieves strong generalization on both classification and segmentation tasks, outperforming prior models on GEO-Bench and Copernicus-Bench.
  arXiv  Detail & Related papers  (2025-06-06T17:59:50Z)
• OmniEarth-Bench: Towards Holistic Evaluation of Earth's Six Spheres and Cross-Spheres Interactions with Multimodal Observational Earth Data [72.98496934729245]
  Existing benchmarks for multimodal learning in Earth science offer limited, siloed coverage of Earth's spheres and their cross-sphere interactions.<n>We introduce textbf OmniEarth-Bench, the first multimodal benchmark that systematically spans all six spheres.<n>Built with a scalable, modular-topology data inference framework and native multi-observation sources, OmniEarth-Bench produces 29,855 standardized, expert-curated annotations.
  arXiv  Detail & Related papers  (2025-05-29T15:02:27Z)
• REOBench: Benchmarking Robustness of Earth Observation Foundation Models [48.24281482353377]
  REOBench is the first comprehensive benchmark for evaluating the robustness of Earth observation foundation models.<n>We conduct a systematic evaluation of a broad range of models trained using masked image modeling, contrastive learning, and vision-language pre-training paradigms.<n>Results reveal that existing Earth observation foundation models experience significant performance degradation when exposed to input corruptions.
  arXiv  Detail & Related papers  (2025-05-22T15:34:50Z)
• DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications [39.0842608370641]
  We present DUNIA, an approach to learn pixel-sized embeddings through cross-modal alignment between images and full-waveform LiDAR data.<n>As the model is trained in a contrastive manner, the embeddings can be directly leveraged in the context of a variety of environmental monitoring tasks in a zero-shot setting.
  arXiv  Detail & Related papers  (2025-02-24T11:28:00Z)
• No Location Left Behind: Measuring and Improving the Fairness of Implicit Representations for Earth Data [13.412573082645096]
  Implicit neural representations (INRs) exhibit growing promise in addressing Earth representation challenges.<n>Existing methods disproportionately prioritize global average performance.<n>We introduce FAIR-Earth: a first-of-its-kind dataset to examine and challenge inequities in Earth representations.
  arXiv  Detail & Related papers  (2025-02-05T16:51:13Z)
• AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalities [5.767156832161819]
  We propose AnySat, a multimodal model based on joint embedding predictive architecture (JEPA) and scale-adaptive spatial encoders.<n>To demonstrate the advantages of this unified approach, we compile GeoPlex, a collection of 5 multimodal datasets with varying characteristics.<n>We then train a single powerful model on these diverse datasets simultaneously.
  arXiv  Detail & Related papers  (2024-12-18T18:11:53Z)
• Embedding Earth: Self-supervised contrastive pre-training for dense land cover classification [61.44538721707377]
  We present Embedding Earth a self-supervised contrastive pre-training method for leveraging the large availability of satellite imagery. We observe significant improvements up to 25% absolute mIoU when pre-trained with our proposed method. We find that learnt features can generalize between disparate regions opening up the possibility of using the proposed pre-training scheme.
  arXiv  Detail & Related papers  (2022-03-11T16:14:14Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.