🤖 AI Summary
To address the limited geographic coverage and scarcity of high-quality labels in global geospatial annotation data, this paper proposes a cross-regional transfer framework based on the AlphaEarth Foundations (AEF) model—marking the first use of its global geospatial representation capability to generalize a fine-grained vegetation classification model (80 classes), trained in the U.S., to Canada. Methodologically, the framework integrates high-density geographic features extracted by AEF with lightweight classifiers—including random forests and logistic regression—thereby circumventing reliance on region-specific annotations. Experiments demonstrate classification accuracies of 81% on the U.S. validation set and 73% on the Canadian validation set. Qualitative analysis confirms strong alignment between predicted labels and ground-truth vegetation distributions. This work significantly enhances the scalability and cross-domain generalization capacity of geospatial models, establishing a reusable technical pathway for remote sensing interpretation in low-resource regions.
📝 Abstract
High-quality labeled geospatial datasets are essential for extracting insights and understanding our planet. Unfortunately, these datasets often do not span the entire globe and are limited to certain geographic regions where data was collected. Google DeepMind's recently released AlphaEarth Foundations (AEF) provides an information-dense global geospatial representation designed to serve as a useful input across a wide gamut of tasks. In this article we propose and evaluate a methodology which leverages AEF to extend geospatial labeled datasets beyond their initial geographic regions. We show that even basic models like random forests or logistic regression can be used to accomplish this task. We investigate a case study of extending LANDFIRE's Existing Vegetation Type (EVT) dataset beyond the USA into Canada at two levels of granularity: EvtPhys (13 classes) and EvtGp (80 classes). Qualitatively, for EvtPhys, model predictions align with ground truth. Trained models achieve 81% and 73% classification accuracy on EvtPhys validation sets in the USA and Canada, despite discussed limitations.