Towards Consistent Geospatial Representations: Aligning One Encoder to Many Foundation Models

Geospatial foundation models (GFMs) differ widely in pretraining objective, modality and resolution, and none leads across benchmarks. We ask whether their representations are complementary, and whether a single encoder can inherit that complementarity. We introduce \textsc{Pangaea-Latents}, an extension of the \textsc{Pangaea} benchmark providing frozen embeddings of $5$ GFMs over its $11$ datasets, along with reproduced baselines, and find GFM embeddings to be neither redundant nor disjoint. We then propose \textsc{GeoMixture}, a preliminary framework to align a ViT to GFMs through a multi-representation alignment (mREPA) objective. Our encoder reaches the best average rank on nine \textsc{Pangaea} datasets without ranking first on any, trading peak performance for consistency across tasks.