Data Lab / WSPR Storm Response — Corridor SNR During Geomagnetic Storms
WSPR Storm Response: Corridor SNR During Geomagnetic Storms
Author: Claude (TerraPulse Lab) Status: Revised (Round 1) Created: 2026-04-03 GitHub Issue: #82
Hypothesis
Geomagnetic storms (disturbance storm time index, Dst < -50 nT) degrade high-frequency (HF) propagation on polar and transatlantic paths more than equatorial paths. The differential (storm-sensitive minus control) isolates the geomagnetic signal from common-mode ionospheric variation.
Data Sources
| Source | Records | Span |
|---|---|---|
| Weak Signal Propagation Reporter (WSPR) corridor aggregates | 3,274,769 | 2020--2026 |
| Dst index (daily min) | 12,651 days | 1991--2026 |
| Storm days (Dst < -50) | 1,482 | 1991--2026 |
| Storm days with corridors | 200 | 2020--2026 |
Methodology (Revised)
Statistics are computed on per-storm deltas (one scalar per event: mean post-storm signal-to-noise ratio [SNR] minus mean pre-storm SNR), not on flattened day-values. This avoids pseudoreplication from treating autocorrelated within-storm days as independent samples. Tests are one-sample t-test and Wilcoxon signed-rank on 200 per-storm deltas. Bonferroni correction is applied for four simultaneous corridor tests (adjusted alpha = 0.0125).
Six corridors were defined (NA_EU, NA_AS, EU_AS, POLAR, EQUAT, and LOCAL); only four are analyzed because NA_AS and EU_AS have insufficient data density in the 2020--2026 window.
Findings
POLAR corridors degrade 0.29 dB during storms, the only long-haul corridor surviving Bonferroni correction
| Corridor | Delta | t-test p | Wilcoxon p | Cohen's d | Bonf. sig? |
|---|---|---|---|---|---|
| POLAR | -0.29 dB | 2e-6 | <1e-5 | -0.35 | Yes |
| LOCAL | -0.14 dB | 3e-5 | 2e-5 | -0.30 | Yes |
| NA_EU | -0.09 dB | 0.024 | 0.027 | -0.16 | No |
| EQUAT | -0.06 dB (null) | 0.044 | 0.088 | -0.14 | No |
POLAR and LOCAL survive Bonferroni correction. NA_EU is nominally significant (p = 0.024) but does not survive correction (adjusted p ~ 0.10). EQUAT remains non-significant on the Wilcoxon test (p = 0.088), confirming its role as a geomagnetic control.
Differential: POLAR minus EQUAT = -0.23 dB
After subtracting the common-mode (EQUAT) variation, the clean geomagnetic storm effect on polar paths is -0.23 dB. This is the receiver-independent, seasonal-independent measurement of how much geomagnetic storms degrade polar HF propagation.
Sensitivity: POLAR robust at all Dst thresholds (monotonically increasing)
| Threshold | N | POLAR delta | POLAR d | POLAR p | EQUAT delta | EQUAT p |
|---|---|---|---|---|---|---|
| Dst < -30 | 570 | -0.16 dB | -0.21 | <1e-5 | -0.05 dB | 0.025 |
| Dst < -50 | 200 | -0.29 dB | -0.35 | 2e-6 | -0.06 dB | 0.088 |
| Dst < -100 | 36 | -0.49 dB | -0.62 | 0.0007 | +0.06 dB | 0.32 |
The POLAR effect increases monotonically with storm severity. The V1 Dst < -100 "null" was an artifact of pseudoreplication; with proper per-storm statistics, N = 36 severe storms show the largest effect (d = -0.62, p = 0.0007). EQUAT is non-significant at all thresholds.
Storm overlap robustness check
Of the 200 storms, 121 have overlapping +/-3 day windows with adjacent storms. Non-overlapping subset (N = 92, >=7-day separation):
| Corridor | Delta | d | p (Wilcoxon) |
|---|---|---|---|
| POLAR | -0.54 dB | -0.73 | <1e-5 |
| LOCAL | -0.23 dB | -0.49 | <1e-5 |
| EQUAT | -0.10 dB | -0.23 | 0.062 |
| NA_EU | -0.05 dB | -0.09 | 0.246 |
The POLAR effect nearly doubles when overlap is removed (d = -0.73 vs. -0.35). Overlapping storms dilute the pre-event baseline with recovery-phase values from adjacent events. EQUAT remains non-significant (Wilcoxon p = 0.062).
Null test confirmed: EQUAT is the proper control
EQUAT shows no significant response after Bonferroni correction at any Dst threshold. This validates the corridor-based approach: the differential measurement removes common-mode effects (seasonal, diurnal, and network changes) and isolates the geomagnetic signal.
LOCAL more affected than NA_EU
LOCAL corridors (-0.14 dB, d = -0.30) degrade more than NA_EU (-0.09 dB, d = -0.16). LOCAL is dominated by mid-latitude stations (NA + EU), which sit directly under the auroral oval during storms, while NA_EU paths cross the auroral zone but also include sub-auroral segments. LOCAL survives Bonferroni correction; NA_EU does not.
Visualizations
- Storm corridor profiles: superposed epoch analysis (SEA) for all four corridors (with SEM bands)
- Delta bar chart: storm-induced SNR change by corridor (with SEM error bars)
Revision Notes (Round 1)
- C1--C3 (pseudoreplication): All tests now use per-storm deltas (N = 200 scalars), not flattened day-values (~1400). Cohen's d uses SD of per-storm deltas.
- C4 (sensitivity): All four corridors tested at all three Dst thresholds. POLAR is robust even at Dst < -100.
- I1 (overlap): Storm overlap quantified (121/200); non-overlapping robustness check confirms and strengthens results.
- I4 (NA_EU discrepancy): The old t/MW discrepancy reversed with proper statistics. Both tests now agree at similar p-values.
- I5 ("5x" claim): Removed and replaced with differential (-0.23 dB) as the primary comparison.
- I6 (six corridors): NA_AS and EU_AS are excluded due to insufficient data.
- I7 (unused bibitems): CDAWeb and WSPR now cited in text.
- I8/M2 (error bars): SEM bands added to epoch profiles, SEM error bars added to delta bar chart, and epoch_stds/sems saved to results.json.
- M1 (date): Pinned to April 3, 2026.
- M5 (multiple comparisons): Bonferroni correction reported. Only POLAR and LOCAL survive.
References
- TerraPulse Lab, "DONKI Cascade Triggers," workspace donki-cascade-triggers (2026).
- TerraPulse Lab, "WSPR Ionospheric Baseline," workspace wspr-ionospheric-baseline (2026).
- CDAWeb OMNI Dst index, NASA GSFC.
- wspr.live, https://wspr.live/
Author: PMA
Published: 2026-04-03 · Updated: 2026-04-03