Geographic substitution of the Uruana residuals | Version 7

Run from the root of this package with Python, NumPy, pandas, SciPy and Matplotlib. This workflow needs no network, new acquisitions or separately installed geomagnetic libraries.
  export OPENBLAS_NUM_THREADS=1
  python background_transplant/transplant.py
  python background_transplant/diagnostics.py
  python background_transplant/verify.py
  python background_transplant/figure.py

Inputs. 2011 uses amas_test/input/flight_2011.csv (SGB project 3087, 348 original 250 m bin medians), using MAGBRU with GPSALT. The 2008 input is nested_gradients/grid2008.csv.gz (ANP 0032, scientific data subset). There are 264 native nodes in the 18 x 32 km bounding footprint, two channels and five corridor half-widths. Both originally observed channels and each bin or grid row are retained. The comparisons share rows, windows and channels. They are not independent replicates.

Transport and scoring. Each geographic coordinate lies on WGS84. Orthogonal projection of its ellipsoidal surface point onto the native center's east and true-north tangent axes defines fixed horizontal offsets. At a candidate center, put these offsets on the new east/north axes and intersect a line parallel to the center normal with the local side of the ellipsoid. Restore each sample's original geodetic height. The operation reproduces native coordinates at Uruana and preserves horizontal components on the local tangent planes; it does not assert equality of all long geodesics. Both years use the same anchor at 16.0593 S, 46.2585 W. Slopes use the unchanged original grid-north coordinate. The fixed part at each sample is measured scalar intensity minus local full-field intensity. Add modeled full-field intensity at the transported sample. Slopes are linear, allowing residual and transplanted background slopes to be added after applying identical weights. 2011 uses OLS by line and window. The 2008 comparison first averages valid nodes in each northing row and then uses equally weighted rows. The WGS84 calculation is independent of the geographic mask and the two periods are 2008.5 and 2011.5.

Synthesis. The large-scale IGRF-14 geocentric north, east and radial components are exact Fourier series through order 13 in geographic longitude. Combine terms at each moved latitude and height before evaluating at the chosen longitudes. Intensity is the norm of the final vector. No spatial interpolation or gradient linearization is applied on the global grids. We use the same published coefficient epochs as the earlier sections. The bundled ppigrf implementation independently recomputes field and slope examples at 19 centers, including the native footprint and Mocuba. Natural Earth supplies the map outline from matched_rotation/inputs/ne_110m_land.geojson. Independent IGRF software, license and source details are in matched_rotation/README.txt.

Definitions. A positive margin means every required signed slope is positive. margin2011 is the least of 28 signed slopes over the original target/north pairs. margin2008 is the least of 40 signed slopes over the 4 bands, 2 channels and 5 widths. Both positive is the two-survey pattern. main2011 and main2008 are these same margins computed using only the transferred background. nested2011 is the least of the 28 full-field slopes and negatives of the 28 degree-one slopes; three_scale adds nested2011>0 to the joint condition. The weak-field definition uses geodetic-surface intensity at the center at BOTH dates. Cells which changed classes are reported separately. For completeness, in the computed grids every site satisfying the more demanding 2011 residual-plus-field criterion also satisfied the transplanted 2008 criterion. Consequently the two tests do not select statistically independent areas. The joint result does not show that 2008 contributed extra spatial discrimination in this calculation.

Results. world_1deg.csv.gz and world_2deg.csv.gz contain geographic center, both surface intensities (nT) and the signed margins (nT/km). same_latitude.csv.gz contains all 1,440 grid centers along the parallel at 0.25-degree spacing. summary.csv gives event counts and area percentages for the entire finite grid, contour class and southern hemisphere; denominators refer to the specified class. The weight column is the sum of cosines of latitude (uniform weights on the single parallel). Multiply by the exact constant strip factor for spherical areas; percentages are unchanged. native.csv and native_constraints.csv retain the original footprints' modeled margins and each individual signed constraint. native_components.csv counts complete original pairs and four-band tests when retaining one or both terms. multiplier_intervals.csv gives open intervals for residual slope plus a times the original main-field slope to preserve all signs. This multiplier scales only spatial variation, not absolute field strength or remanent magnetization. longitude_runs.csv reports first and last eligible cell centers, not boundary estimates. latitude_bands.csv lists follow-up comparisons across latitude bands. verification.json records numerical cross-checks. Synthesis and input files are fully offline.

Mocuba is a post hoc geographic example chosen after the exterior band near southeastern Africa was found. The station coordinates 16°50′ S, 36°59′ E are from the station list, no. 508, Annex 2 of “Evaluation of length of growing periods and crop growing possibilities in Mozambique”, https://edepot.wur.nl/488333. mocuba.csv evaluates the two modeled regional backgrounds at that center. The survey residuals, heights and geometries are entirely Brazilian. No measured Mozambican magnetic anomaly or repeated local survey is asserted. These tests do not transport the source magnetization, distinguish the source mineralogy, or attribute causality for the South Atlantic Anomaly. Percentages refer to compatibility with this one template on a modeled grid rather than the occurrence of natural anomalies.
