Provenance-audited resampling of underwater acoustic delay statistics from Korean coastal measurements

Archived underwater acoustic channel statistics are attractive simulation inputs, but their provenance and processing are rarely verifiable. This paper tests a procedure for reusing archived delay statistics, demonstrated on workbooks from three Korean coastal sites, one serving only as a provenance and sparse-support case. The procedure audits environment provenance against the acquisition archive, re-derives archived statistics from surviving raw waveforms, identifies algebraically derived columns, and resamples delay statistics only within identifiable, support-graded conditions under acquisition-blocked validation. Of 400 East and South Sea records, 319 were laboratory, lake, or dockside tests; the West Sea workbooks listed nearly every frame three times. Re-deriving 441 raw acquisitions reproduced the archived estimator, a matched filter over a 50 ms window. It also showed that 21% of West Sea frames were timed on noise or noise-limited, producing apparent delays up to the window bound, and that one link carried a fixed 11.4 ms offset. The archived coherence bandwidths were rule-of-thumb transforms of the RMS delay spread. On the identifiable data, within-condition resampling error is set by sample size, whereas no tested route (pooling, kernel smoothing, or parametric or physics-based models) transfers reliably to held-out distances, links, days, or campaigns, so out-of-support queries are refused by default. On public Watermark channels, row-level splits understated error by median factors up to 7.6; a design-effect approximation ranked them. The results describe the surveyed site–campaign configurations, not sea regions. The resampler generates scalar delay statistics, not time-varying impulse responses; the audited version 2.0 data and code accompany the revision.