HISTORICAL DEPTH/DAMPING SENSITIVITY — 5 November 2026

Online Resource 1 for:
The Uruana de Minas magnetic anomaly within the South Atlantic Anomaly in Brazil
Lucas Giovani Ribeiro and Marcos André Simonssini
Independent researchers, Uruana de Minas, Minas Gerais, Brazil
Corresponding author: Lucas Giovani Ribeiro, coin26@mail.com

Purpose
Assess whether the nearby historical profile-shape improvement is confined to
the primary equivalent-source depth and ridge damping. This is an additional
post hoc sensitivity analysis, designed after the primary result was known.
It is not a new independent validation sample or a parameter-selection step.

Design
Use the exact same 75 local segments (608 points) and IBGE-corrected coordinates
as distant_transfer. Source coefficients are fitted to 2008 data only.
Vary source depth below the nominal 2008 flight plane over 4, 6, 9 km and
the ratio of ridge damping to the largest singular value over 0.01, 0.03, 0.10.
These nine combinations are the same depth/damping grid used by the existing
orbital sensitivity test; this audit uses the diurnally corrected channel.
Retain the source spacing, training disks, survey heights, 1971.5 field epoch,
segment membership and within-segment centered, point-weighted RMSE.
Compare frozen, projection-only and full epoch-update predictions.
No configuration is selected from the historical evaluation errors.

Files and reproduction
run.py         Recompute all 675 segment/configuration results and summaries.
segments.csv   Per-segment sums of squared centered errors for all models.
summary.csv    Nine pooled scores and segment/block improvement counts.

From the extracted Online Resource 1 root:
OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 python historical_sensitivity/run.py
Dependencies are NumPy, SciPy and pandas, as listed in the root README.
The script imports the existing coordinate transformation, field model and
dipole kernel. It reads only the archived inputs. It checks the primary
configuration against the independently archived distant_transfer score
with absolute RMSE tolerance 1e-9 nT.

Results
Full-update gain over the frozen field is positive in all nine configurations:
10.6270% to 35.2396%. The primary 6 km / 0.03 configuration reproduces the
archived frozen and full-update RMSEs, 2.658931 and 1.805546 nT (32.0950%).
Between 49 and 59 of the 75 segments improve, and 5–7 of 9 geographic blocks.
Projection alone outperforms the full update in five configurations. The
benefit of changing field epoch persists on this grid; the gain magnitude
and attribution to an induced-moment update depend on source parameterization.
This audit does not extend the depth/damping analysis to the distant or
Alaskan samples and does not identify physical source depths.
