Scaling sparse Earth-observation measurements to detailed maps

Develop accurate and efficient methods for scaling highly detailed but spatially and temporally sparse ground-based labels and measurements into maps at global scales.

Background

Global mapping applications often depend on high-quality labels and field measurements that are spatially localized and temporally sparse. Although Earth-observation archives provide extensive coverage, translating these detailed but unevenly distributed observations into accurate maps remains difficult, particularly when measurement precision and spatial coverage must be balanced.

The paper presents AlphaEarth Foundations as a learned geospatial representation intended to extrapolate sparse labels and measurements. The authors nevertheless identify the broader problem of accurately and efficiently scaling such observations as unresolved, motivating continued methodological development beyond the reported model and evaluations.

References

Accurate and efficient scaling of such highly-detailed yet spatially and temporally sparse data remains an open challenge, e.g., \citep{sun2021research}.

— AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data  (2507.22291 - Brown et al., 29 Jul 2025) in Section 2, “From sparse labels to maps”