Internal analysis design recorded before computing the following new comparisons. The Uruana 2011 target and prior regional-model results are already known. This procedure is retrospective.
1. On the original 2011 per-line 250 m bin medians, compute model intensity for degree 1 only and full IGRF-14 at exactly the same track points and GPS elevations. Fit an OLS slope vs grid north in each of the 14 pre-existing target windows and corresponding northern return windows. Test the sign sequence dipole / full / measured (-,+,- for target, -,+,+ for north). Test the within-line target and north together. Summarize median and range; shared lines are not independent replicate regions. Also compare the supplier IGRF column, and both reference fields at constant survey median elevation.
2. Keep the measured scalar residual relative to full IGRF fixed and substitute the degree-1 background on the same route. Count slope signs on target and northern return. A change in classified gradient or in sign recovery is arithmetic conditional on the fixed residual and does not model how the crust would remagnetize. Degree one is the actual tilted centered dipole at the date. Increasing degree truncations 1 to 13 give a descriptive sensitivity, not a unique-source decomposition. Record the non-dipolar increment Ffull minus F1 as an intensity difference.
3. Estimate the horizontal true-north derivative at the survey center independently of the flight tracks (centered differences ±1 km and fixed altitude). Compare degree 1 and full model every degree in longitude at Uruana latitude at 2011.5. Report the sign reversal inside and outside the surface F<28 µT contour. Repeat at other survey dates and at adjacent latitudes. Provide an area-weighted southern-hemisphere diagnostic separated by magnetic latitude bands if useful, explicitly distinguish descriptions of a modeled field from evidence of local crustal gradients. No global rarity or priority inference.
4. Test the second reversal across native data controls rather than counting synthetic local anomalies. Construct non-overlapping sets of 14 consecutive 2011 line identifiers aligned to the original group, reproduce the target's geometry with the disk cap replicated at each candidate group midpoint, use the two existing northern/southern control corridors and common survey eligibility, at least 4 250m bins in target and 8 in north, comparable Y span. For each full-coverage group count line pairs satisfying BOTH measured negative in cap and positive in recovery where dipole is negative and full model positive in both windows. Score target and controls identically; report tied or stronger controls and account for nesting. If previously aggregated bins differ, re-extract from project 3087 or qualify.
For the regional control extraction use groups −8 through +8 at 7 km center spacing (fourteen nominal 0.5 km line spacings), each inside the preceding regional footprint of ±65 km east. Use north offsets −32, 0 and +32 km. Require within-line spans of 0.75 and 2 km for target and north respectively, and report complete 14-line groups separately. Check the compensated channel as a sensitivity. Degree-by-degree subtraction from the full vector model, if performed, is a follow-up to inspecting the truncation sequence.
Follow-up after seeing the 2011 control counts, before inspecting 2008 at these coordinates. Apply the PRE-EXISTING 2008 four-band (0–8, 8–16, 16–24, 24–32 km) scalar test to all 51 nominal sector origins, not just those passing in 2011. Both 2008 channels and all five prior corridor half-widths (0.75, 1.5, 3, 6, 9 km) must reproduce + − + +. Each band/width/channel requires at least four occupied northing rows and 80% of the possible native nodes. Use the established 2008 center (365381.662,8224096.138 m) plus the same group displacements; also translate by 28 m to the 2011 native origin as sensitivity. Recalculate the four-band test after replacing the scalar full-field trend with the degree-one trend while keeping observed residual fixed, using 2008.5 and the 2.2 km flight plane. Report the cross-survey denominator, both candidate sites even if support fails, and results in both backgrounds. This prospective-in-the-workflow test is a retrospective application of a known criterion, not an independent discovery of Uruana.
Additional sensitivity chosen after the controls. Add the same uniform slope s (nT/km of grid north) to all observed line and band slopes. The bounds for preserving every sign are obtained algebraically. For 2011 require s > minus the smallest northern slope and s < minus the largest target slope. For the 2008 four-band rule use the same expressions for its positive and negative classes, and intersect the two open intervals. This describes signal margins with the observed data fixed, without an uncertainty or AMAS causality interpretation.
