Search papers, labs, and topics across Lattice.
This study investigates the temporal sensitivity of Tessera embeddings for land-use/land-cover mapping by analyzing how varying observation windows鈥攆rom a full year to a single day鈥攁ffect performance. By keeping the encoder frozen and benchmarking against both linear probes and UNet segmentation heads, the authors reveal that the effectiveness of embeddings is highly task-dependent, with significant performance variations based on class phenology and temporal stability. Key findings indicate that while shorter temporal windows lead to gradual degradation in accuracy, they still allow for effective classification, suggesting that temporal coverage can be optimized for specific applications in Earth Observation.
Temporal coverage in Earth Observation embeddings is a tunable cost, enabling near-real-time mapping without sacrificing accuracy for certain land-use classes.
Many Earth Observation applications need land-use/land-cover maps that are both precise and frequently updated, yet the strongest Earth Observation foundation models build their embeddings from a full year of observations. We present a controlled study of the temporal sensitivity of Tessera, one of these leading foundation models, for land-use/land-cover mapping. Keeping the encoder frozen, we recompute its embeddings over varying observation windows, from a full year down to a single day. We use them as inputs to a linear probe and a UNet segmentation head, benchmarking both of them against from-scratch networks on LUCAS, DynamicEarthNet, and PASTIS-R datasets. We show that the value of the embeddings is task-dependent. Where classes are separated by phenology, as for the crop types of PASTIS-R, they reach a mean Intersection-over-Union of $58.3$, about $46\%$ above the best from-scratch model. Where classes are temporally stable (e.g., forests in DynamicEarthNet and LUCAS), embedding-based and from-scratch models match only under full supervision. On both datasets, Tessera embeddings remain markedly more label-efficient. Degradation under shorter temporal windows is gradual and class-dependent. Contracting the window from one year to one month costs $39\%$ of the segmentation accuracy on PASTIS-R but only $5\%$ on DynamicEarthNet. Single-day embeddings still classify land cover in LUCAS at $3.4$ times the chance level. Our study shows that temporal coverage is therefore a tunable cost rather than a fixed prerequisite, opening regimes such as near-real-time mapping and faster land-use/land-cover refresh cycles.
