The presence and duration of water in wetlands drive core biogeochemical processes that shape ecosystem function and carbon cycling. Yet data on wetland inundation regimes remains limited beyond static inventory maps, especially in data‑scarce regions. This gap is compounded by the computational and storage demands of deriving inundation information from dense satellite time series, particularly for small, cryptic wetlands that are ephemerally flooded or masked by vegetation. Emerging Earth observation foundation models, such as Google’s AlphaEarth, compress spectral, temporal, and contextual information from petabytes of multi‑source observations into generalizable geospatial embeddings that offer a novel approach for monitoring land surface processes. In this study, we use a data‑rich wetland site in the Florida Everglades to evaluate how AlphaEarth embeddings could support inundation mapping in both data‑rich and data‑scarce regions. We first assess embeddings in a supervised setting by training a Random Forest classifier to map annual inundation regimes (e.g., seasonal, permanent, ephemeral) and their trajectories from 2018–2021 at 10 m resolution, using AlphaEarth embeddings and ground‑truth data from the Everglades Depth Estimation Network (EDEN). We compare accuracy to the same model trained on annual Sentinel‑2 Normalized Difference Water Index (NDWI) composites to quantify the added value of embeddings over raw satellite data. Next, to represent sites lacking ground truth data, we assess embeddings in an unsupervised setting by clustering in embedding space and comparing these clusters to those derived from the Sentinel‑2 NDWI composites. In this unsupervised case, EDEN data are used solely to evaluate how well clusters correspond to distinct inundation regimes, not for training. Finally, we benchmark both supervised and unsupervised results against inundation regimes reported by the 30 m Global Surface Water Explorer and the National Wetlands Inventory. Together, these analyses evaluate the potential of AlphaEarth embeddings to advance diverse applications of wetland remote sensing. The resulting maps can help pinpoint biogeochemical hotspots, track regime changes through time, inform the parameterization and fine-tuning of foundational and process-based models, and guide targeted field sampling campaigns.