Inferring Height from Earth Embeddings: First insights using Google AlphaEarth

2026-02-19

AI Summary

This study investigates the decoding of surface elevation information from 10-meter-resolution Earth embeddings provided by Google AlphaEarth to enable high-precision regional elevation modeling. We present the first empirical validation that AlphaEarth embeddings contain transferable topographic signals, leveraging lightweight U-Net and U-Net++ convolutional decoders for height regression, supervised by high-accuracy digital surface models (DSMs). Experimental results demonstrate that U-Net++ significantly outperforms U-Net in generalization under distribution shift, achieving a test-phase R² of 0.84 (versus 0.78 for U-Net) and a median error of only −2.62 meters (compared to −7.22 meters for U-Net), with a training R² as high as 0.97. These findings confirm the strong terrain-related content embedded in AlphaEarth representations and highlight that mitigating regional bias is crucial for further performance gains.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Representation Learning for Vision
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This study investigates whether the geospatial and multimodal features encoded in \textit{Earth Embeddings} can effectively guide deep learning (DL) regression models for regional surface height mapping. In particular, we focused on AlphaEarth Embeddings at 10 m spatial resolution and evaluated their capability to support terrain height inference using a high-quality Digital Surface Model (DSM) as reference. U-Net and U-Net++ architectures were thus employed as lightweight convolutional decoders to assess how well the geospatial information distilled in the embeddings can be translated into accurate surface height estimates. Both architectures achieved strong training performance (both with $R^2 = 0.97$), confirming that the embeddings encode informative and decodable height-related signals. On the test set, performance decreased due to distribution shifts in height frequency between training and testing areas. Nevertheless, U-Net++ shows better generalization ($R^2 = 0.84$, median difference = -2.62 m) compared with the standard U-Net ($R^2 = 0.78$, median difference = -7.22 m), suggesting enhanced robustness to distribution mismatch. While the testing RMSE (approximately 16 m for U-Net++) and residual bias highlight remaining challenges in generalization, strong correlations indicate that the embeddings capture transferable topographic patterns. Overall, the results demonstrate the promising potential of AlphaEarth Embeddings to guide DL-based height mapping workflows, particularly when combined with spatially aware convolutional architectures, while emphasizing the need to address bias for improved regional transferability.

Problem

• Earth Embeddings
• surface height inference
• generalization
• distribution shift
• Digital Surface Model

Innovation

• Earth Embeddings
• AlphaEarth
• surface height inference
• U-Net++
• geospatial deep learning