Northern recovery in the 2011 Uruana profiles
This retrospective sensitivity test follows the earlier finding that subtracting the local IGRF-14 intensity removes northern recovery on six raw profiles under ordinary least squares. All 14 native target and northern windows, archived products and slope estimators were retained. No window was moved or omitted to select the eastern lines. Consecutive line numbers run west to east in the Uruana data.

Run from the extracted root:
  python local_recovery/recovery.py
  python local_recovery/figure.py

line_slopes.csv reports measured total-field, IGRF-14 and residual slopes (nT/km) for each line, window, channel and estimator. The modeled main-field intensity is evaluated at 2011.5 at each bin's coordinates and GPS altitude before subtraction. Both slopes are fitted to the values at the same grid northings. For Theil–Sen, the slope of the residual is estimated from the residual profile itself. The code checks the 280 measured slopes against processing_audit and the 28 raw least-squares decompositions against background_transplant.

line_recovery.csv records, for each line and analysis, whether the northern window has a positive measured slope and a nonpositive residual slope. The five lines 12430, 12440, 12450, 12460 and 12470 always change sign, strictly from positive to negative on subtraction. Line 12380 meets this criterion under raw and compensated ordinary least squares only. The other eight lines have a positive northern residual with full bins under every configuration.

omit_one_north.csv reports both slopes after dropping each individual occupied 250 m bin, retaining a minimum of eight bins and 2 km span. The five eastern tracks each allow 17 omissions. All 85 retain a positive observed and a negative residual slope under every configuration. Other lines are reported as well. Survey bins, lines and processing configurations are dependent.

Figure_02.pdf and Figure_02.png are generated from local_recovery/line_slopes.csv, line_recovery.csv and regional_specificity/results/sectors_1deg.csv. Panel (b) uses the maximum raw OLS margin for each sector, evaluated over the global 1° grid. Its points represent measured Brazilian sectors; the candidate main fields are modeled.
