Seismic processing on obsidian (NPRA line 31-81) - prediction log.
Author: Claude Opus 5.5 (claude-opus-5-5). Series letter: E.
Written BEFORE opening any of the line's files, logs or the SEG wiki page about it. Already seen:
the archive index (file names, shot points 100-155, sizes 35.37 + 35.37 + 9.44 MB raw, 3.18 MB
processed stack) and a search snippet mentioning an NPRA "seismic shothole sample" dataset.
Format: ID [confidence] claim. See ../../PREDICTIONS-GUIDE.md.

--- round 1: the data ---
E01 [0.75] The textual (3200-byte) headers of the raw files are EBCDIC, not ASCII.
E02 [0.45] Raw samples are IBM System/360 floating point (format code 1).
E03 [0.75] The source is dynamite in shot holes, not Vibroseis.
E04 [0.50] Sample interval is 2 ms (rather than 4 ms or 1 ms).
E05 [0.40] Record length is between 5 and 6 s inclusive.
E06 [0.45] Each field record has between 96 and 120 data channels inclusive.
E07 [0.55] The observer's log is handwritten or typed on a printed form (scanned), not machine output.
E08 [0.60] The 1981 processed stack (31_81_PR.SGY) has 4 ms samples (resampled from the field rate).
E09 [0.50] The processed stack has at least 24-fold coverage as stated in its headers or the logs.

--- round 1: the machine ---
E10 [0.50] Seismic Unix (current release) builds on AIX 4.3.3 with gcc 2.95.3, needing changes to
           fewer than 5 source files.
E11 [0.65] IBM-float to IEEE conversion on the 604e runs faster than 5 million samples/s in C.
E12 [0.60] A full flow on the line (gain, decon, CMP sort, NMO, stack, post-stack migration) takes
           under 10 minutes of wall time on obsidian.
E13 [0.50] Migration is the most expensive stage of that flow.
E14 [0.80] Our brute stack shows the same main reflectors as the 1981 stack (judged by eye).
Sat Oct 17 01:42:09 AM MDT 2026

=== OUTCOMES (round 1, data part; scored 2026-10-17) ===
E01 HIT      [0.75] textual headers of all three raw files and the stack are EBCDIC (USGS 2023 header text on the raw files; the original 1981 header on the stack).
E02 HIT      [0.45] format code 1, IBM floating point, in raw files and stack.
E03 HIT      [0.75] "ENERGYSOURCE DYNAMITE" (raw textual header); the observer's log has charge size and hole depth columns.
E04 HIT      [0.50] 2000 us in binary and trace headers; "SAMPLERATE 2".
E05 HIT      [0.40] 6000 ms (3000 samples at 2 ms).
E06 HIT      [0.45] 96 channels per the header (files hold 101 traces/record: 100 data + 1 aux in the binary header; to be sorted out).
E07 HIT      [0.55] handwritten on a printed GSI "Land Seismic Recording Log" form, scanned (party 1186, 12 March 1981).
E08 HIT      [0.60] stack 31_81_PR.SGY: 4000 us, 1501 samples (6 s).
E09 MISS     [0.50] fold 12 ("CHANNELS 96 FOLD 12 GROUPINTERVAL 110"; 96 x 110 ft / (2 x 440 ft) = 12).
Not predicted: the 1981 stack was processed on a Texas Instruments ASC ("INSTRUMENT: MFG TI MODEL ASC",
"FIELD TAPE PROCESSING MACHINE NUMBER IS: ASC"), not on IBM; the crew was GSI's (Geophysical Service Inc., TI's seismic subsidiary; TI grew out of GSI).
The raw trace headers carry no geometry at all (offset, sx, gx, ep, cdp all 0): it must come from the
observer's log. File 1 holds 30 field files for 20 shot points (tests and re-shoots).

--- round 2, after reading the headers and the observer's log, before processing any traces ---
E15 [0.70] Traces 97-101 of each record are auxiliary (not reflection data); 1-96 are the groups.
E16 [0.55] Group 1 is at the low-station (north) end, so x = 440*SP + 110*(g-48.5) is right as
           written (test: the stack with this geometry is visibly better than with groups reversed).
E17 [0.60] First breaks show a near-surface velocity between 5,000 and 8,000 ft/s at far offsets
           (permafrost: fast).
E18 [0.50] The raw samples carry recording gain: amplitudes span more than 4 orders of
           magnitude between early and late times on a near trace.
E19 [0.65] SP 110 and SP 143 ("vented") are visibly noisier or weaker than their neighbours.
E20 [0.75] The 1981 stack has one trace per 55 ft CDP across the whole midpoint range, taper
           included: shots span 55 x 440 = 24,200 ft, midpoints reach 2,612 ft beyond each end shot,
           so about 29,425 / 55 = 535 CDPs, against the 534 traces in the file.
           (E20 rewritten at 02:33 before any new look: the first draft was arithmetic, not a claim.
           Caveat: segyhdr.py output seen earlier showed the stack's fldr field in groups of 8 traces,
           which already hints at 8 CDPs per 440 ft shot interval. Not fully independent.)
Sat Oct 17 02:31:15 AM MDT 2026

=== OUTCOMES (round 2, part; scored 2026-10-17 from a quick look in numpy on kootenai, not yet on obsidian) ===
E15 HIT      [0.70] traces 97-101 hold a near-constant ~1e9 level from 0 to 6 s (no reflections); 1-96 are the groups.
E17 MISS     [0.60] first breaks on file 120 (SP 109): 10,500-13,700 ft/s at offsets beyond 2,500 ft and
                    about 12,000 ft/s inside 1,500 ft: much faster than I guessed (frozen ground).
E18 MISS     [0.50] near trace 49, file 120: mean |x| 1.06e9 at 0.05-0.2 s down to 2.6e5 at 5-6 s: 3.6
                    orders, and the last 2 s are a flat noise floor. Near traces clip at 2^31 in the first 0.2 s.
E19 HIT      [0.65] 0.5-3 s rms over 5-6 s rms: SP 110 0.9 (neighbours 5-67), late noise 8.7e6 against ~2e5;
                    SP 143 14.6 with late noise 3.8e6 (neighbours 2-9e5). The misfire (file 157) is all noise.
E16 and E20 need the stack.

=== OUTCOMES (round 1 machine part + rest of round 2; scored 2026-10-17 on obsidian) ===
E10 PARTIAL  [0.50] the libraries + 22 programs the flow needs built after changing 1 file (tempname.c,
                    a C99 mid-block declaration); 5 more libsu files fail the same way (unused here), so a
                    full SU build would need at least 6 files changed. A full build was not attempted.
E11 HIT      [0.65] ibmbench.c: 5.8 M samples/s (gcc -O2, bit manipulation), 6.5 (xlc -O3); 1.3 via ldexp().
E12 PARTIAL  [0.60] with Stolt migration the compute stages total about 7 min (read 90 s, preprocessing
                    ~70 s, sort 9, velocity analysis 130, NMO+stack 24, edit+restack 42, Stolt 33);
                    with phase-shift migration (868 s) about 21 min, with Kirchhoff (274 s) about 11 min.
E13 PARTIAL  [0.50] depends on the method: Stolt (33 s) is cheaper than decon (38 s) and velocity analysis
                    (130 s); phase-shift (868 s) and Kirchhoff (274 s) are by far the most expensive stages.
E14 HIT      [0.80] same main reflectors as the 1981 stack, including the strong event near 2.8 s and the
                    dipping package near 1.2 s.
E16 HIT      [0.55] adjacent-trace correlation 0.776 (as written) vs 0.736 (reversed); the reversed stack
                    visibly smears the dipping beds near 1.2 s.
E20 HIT      [0.75] 534 traces, CDP 101-634 in steps of 1; our 55 ft bins give 536 (1-536) for the same line.
                    (Not fully independent, see the caveat on E20.)

Calibration (E01-E20):  <0.50: 3/3   0.50-0.79: 10/16 (3 partial, 3 miss)   >=0.80: 1/1
Not predicted: gcc 2.95 needs ~10 s per SU source file, so a full build is hours; phase-shift migration
is 26x slower than Stolt here; velocity analysis, not migration, dominates a Stolt flow.