Determine whether broader stratum coverage improves downstream performance

Determine whether the broader class–season–region stratum coverage achieved by STRATA-S12 improves downstream performance relative to less balanced SAR–optical corpora through a controlled cross-corpus comparison with geographically held-out evaluation.

Background

The paper validates STRATA-S12 with a pre-registered frozen linear-probe experiment, but this experiment only tests whether a backbone pretrained on a representative subset of the corpus carries transferable signal compared with random initialization. It does not compare STRATA-S12 against other corpora or isolate the causal contribution of broader geographic and stratum coverage.

The authors state that answering this question requires a controlled cross-corpus evaluation using geographically held-out data. This remains unresolved because the reported metadata-level coverage comparison is descriptive and does not measure downstream representation or task performance.

References

We also stop short of one neighbouring claim: we do not test here whether the broader stratum coverage reported in Section~\ref{sec:coverage-comparison} improves downstream performance, since that benefit-of-coverage question requires a controlled cross-corpus comparison with geographically held-out evaluation, which we leave to the follow-up work.

— Deficit-Driven Stratified Acquisition for Building STRATA-S12: An Open SAR-Optical Corpus for Self-Supervised Learning  (2610.17680 - Sanchez-Leon et al., 13 Oct 2026) in Section 5, Technical Validation, opening paragraph (Section 5.1 context)