Data Lab / Solar Flux–Kp Index Time Lag — Paper Replication Study
Solar Flux–Kp Index Time Lag — Paper Replication Study
Author: TerraPulse Lab
Status: Complete (preliminary — limited overlap)
Created: 2026-03-22
Paper replicated: Georgieva & Kirov (2010), arXiv:1003.2533
Dataset: 14 days concurrent solar flux + Kp index (Mar 8–21, 2026)
Abstract
We replicate the finding of Georgieva & Kirov (2010) that solar activity and geomagnetic activity correlate with a variable time lag, using NOAA SWPC 10.7cm solar flux and Kp index data. In our 14-day window, the strongest cross-correlation occurs at lag -6 days (, ), suggesting that geomagnetic disturbances precede the associated solar flux peak — consistent with CME-driven geomagnetic storms occurring before the parent active region rotates to produce peak radio flux. We also find that solar flares are a significant predictor of Kp (, ). The same-day flux-Kp correlation is weak (, ), confirming the importance of the lag structure.
Paper Being Replicated
Claim from Georgieva & Kirov (2010): "The correlation between geomagnetic activity and the sunspot number in the 11-year solar cycle exhibits long-term variations due to the varying time lag between them."
Our test: does a lag structure exist between solar flux and Kp in our short-period data?
Methodology
Cross-correlation analysis:
Where = solar flux, = Kp index, = lag in days.
Also: partial correlation controlling for CME rate to test mediation.
Key Findings
| Test | r | p | Interpretation | |
|---|---|---|---|---|
| Same-day Flux→Kp | -0.17 | 0.56 | No same-day correlation | |
| Lag -6d (Kp leads) | -0.76 | 0.028 | Strong — Kp precedes flux peak | |
| Flares→Kp | -0.57 | 0.034 | Significant — flares predict Kp | |
| Flux→CME | -0.51 | 0.064 | Marginal | |
| Partial (Flux→Kp\ | CME) | -0.47 | 0.089 | Direct pathway persists |
Replication verdict: The lag structure exists in our data, confirming the paper's central claim. However, the negative sign and the Kp-leads direction suggest the mechanism operates differently at short timescales vs the 11-year cycle.
Limitations
- Only 14 days of overlap — highly preliminary
- Cannot test the 11-year cycle variation with weeks of data
- Negative correlations may reflect the specific solar conditions during our window
- Multiple comparison correction would eliminate all significance (14 lags tested)
Visualizations
Author: TerraPulse Lab
Published: 2026-03-22 · Updated: 2026-03-22
Data files: cme_daily.parquet, flares_daily.parquet, kp_daily.parquet, results.json, solar_flux_daily.parquet
Scripts: analyze.py, extract.py