ZTF Optical Transients vs Space Weather: A Null Result

Author: Claude (TerraPulse Lab)

Status: Complete

Created: 2026-04-05

GitHub Issue: #77

Hypothesis

H0 (null): The detection rate of optical transients from the Zwicky Transient Facility (ZTF) is independent of space weather conditions (geomagnetic Kp index, solar wind speed, and solar X-ray flux).

H1 (alternative): The rate of unclassified ("unknown") transients increases during geomagnetic storms, due to atmospheric scintillation degrading classification accuracy.

H2 (geometric): Solar system object detection rate varies independently of space weather (geometric effect).

Data Sources

Metric	Records	Date Range	Notes
`ztf_sn_candidate`	1,905	2026-03-24 to 2026-04-05	ZTF supernova candidates (magnitude)
`ztf_kilonova_candidate`	517	2026-02-09 to 2026-04-02	ZTF kilonova candidates
`ztf_solar_system_mpc`	1,612	2026-03-24 to 2026-04-05	MPC-confirmed solar system objects
`ztf_solar_system_candidate`	1,084	2026-03-24 to 2026-04-05	Solar system candidates
`ztf_unknown`	1,350	2026-03-24 to 2026-04-05	Unclassified transients
`space_kp_index`	198 days	2025-09-15 to 2026-03-31	Planetary Kp (0-9 scale)
`solar_wind_speed`	132 days	2025-10-10 to 2026-04-05	DSCOVR solar wind (km/s)
`solar_xray_flux`	20 days	2026-03-17 to 2026-04-05	GOES X-ray flux (W/m2)

Total overlap: 20 days for ZTF-Kp, 13 days for ZTF-wind/X-ray.

Methodology

Daily aggregation: Count ZTF detections per day per category. Compute daily max Kp, average solar wind speed, and max X-ray flux.
Correlation: Pearson and Spearman correlations between daily detection counts and each space weather metric.
Group comparison: Welch t-test and Mann-Whitney U test comparing detection rates on high-Kp (>=4) versus low-Kp (<4) days, reporting Cohen's d effect size.
Chi-squared test: Test whether the category distribution (SN/KN/solar system/unknown) shifts between high-Kp and low-Kp days.
Sensitivity analysis: Repeat group comparison at Kp thresholds of 3, 4, and 5.
Confound check: Determine whether significant results reflect artifacts of unequal temporal coverage between ZTF categories.

Findings

Primary result: Detection rate is independent of space weather (NULL)

Test	r	p	N	Significant?
Total detections vs max Kp	-0.007	0.977	20	No
Total detections vs avg wind speed	0.204	0.505	13	No
Total detections vs log10(max X-ray)	-0.149	0.628	13	No
Kilonova count vs max Kp	0.274	0.304	16	No

None of these correlations approach statistical significance. The Kp correlation (r=-0.007) is essentially zero. ZTF detection rates show no relationship to geomagnetic or solar activity.

Group comparison: High-Kp vs Low-Kp days

High-Kp days (max Kp >= 4): N=9, mean=166.0 detections/day
Low-Kp days (max Kp < 4): N=11, mean=395.9 detections/day
Welch t-test: t=-0.79, p=0.44
Mann-Whitney U: p=0.13
Cohen's d = -0.33 (small effect, not significant)
Rate ratio = 0.42

The apparent lower rate on high-Kp days is not significant and reverses at different thresholds (see sensitivity analysis), confirming it is noise.

Sensitivity analysis (Kp threshold)

Threshold	N_high	N_low	Ratio	Cohen's d	t-test p	MW p
Kp >= 3	16	4	6.58	+0.43	0.141	0.354
Kp >= 4	9	11	0.42	-0.33	0.443	0.126
Kp >= 5	5	15	0.86	-0.06	0.883	0.499

The direction and magnitude of the "effect" flip depending on the threshold, confirming no real relationship.

Chi-squared distribution shift: Confounded artifact (confirmed by restricted test)

The full-window chi-squared test (chi2=712, p<0.001, Cramer's V=0.35) found a significant difference in category proportions between high-Kp and low-Kp days. However, this is an artifact of unequal temporal coverage:

Kilonova data starts on 2026-02-09; other categories start on 2026-03-24.
The 12 kilonova-only days (Feb-Mar) fell in a mix of high-Kp and low-Kp periods.
When kilonova is the only category on a given day, it dominates that day's distribution.
The "shift" is a sampling artifact from pipeline start dates, not a space weather effect.

Restricted test (revision): Restricting to the common-coverage window (>= 2026-03-24, eight days) where all five pipelines were active simultaneously yields chi2=4.3, p=0.37, Cramer's V=0.03. No significant distribution shift remains, and the artifact disappears.

Magnitude vs Kp: confound confirmed (r=0.66 collapses)

Full-window correlation: r=0.66, p=0.002, N=20. Originally flagged for V2 verification.

Restricted test (revision): Restricting to the common-coverage window (>= 2026-03-24, N=8): Pearson r=0.63, p=0.09 (loses significance at alpha=0.05). Spearman rho=0.75, p=0.03 (nominally significant but N=8 falls below the threshold for reliable inference). The temporal confound explanation holds: early-period kilonova-only detections had a different magnitude distribution, and the early and late periods had different Kp values. V2 flag downgraded to a note.

Per-category tests: four of five untestable

We could not perform per-category correlation tests for supernovae, both solar system categories, and unclassified transients individually (N=8 for each, below the N>=10 threshold). Only kilonova had sufficient coverage (N=16, r=0.27, p=0.30). The aggregate null conclusion rests primarily on the total-detection test.

Power statement

At N=20, a two-tailed Pearson test has 80% power to detect |r| >= 0.57. Effects in the moderate range (|r| ~ 0.2-0.5) remain unresolvable with the current baseline.

Null result (Hypothesis 2): Unknown fraction vs Kp

Only eight overlapping days had both ztf_unknown data and Kp data, which is insufficient for analysis. The hypothesis that atmospheric scintillation during storms degrades classification (increasing the unknown fraction) could not be tested.

Discussion

This is the expected null result. Ground-based optical telescopes like ZTF operate in the visible wavelength band (400-700 nm), which geomagnetic activity does not affect. Space weather primarily affects:

The ionosphere (radio frequencies, WSPR, GPS)
Charged particle environments (satellites, polar aircraft routes)
Power grids (geomagnetically induced currents)

Space weather does not affect optical seeing, sky transparency, or CCD sensitivity. The atmosphere is opaque to the charged particles and X-rays that characterize space weather. Tropospheric weather (clouds, humidity, and wind), light pollution, and lunar phase affect optical astronomy, and none of these correlate with geomagnetic activity.

The 24-day baseline is the principal limitation. While these results point toward a null, a definitive demonstration would require months to years of continuous coverage spanning multiple solar activity phases.

References

Bellm et al. (2019), "The Zwicky Transient Facility: System Overview," PASP 131, 018002
Masci et al. (2019), "The Zwicky Transient Facility: Data Processing, Products, and Archive," PASP 131, 018003
Bartels, J. (1949), "The standardized index Ks and the planetary index Kp," IATME Bulletin 12b
NOAA Space Weather Prediction Center, https://www.swpc.noaa.gov/
Prior TerraPulse workspace: workspaces/geomagnetic-storm-cascades/ (Kp cascade analysis)
Prior TerraPulse workspace: workspaces/solar-terrestrial-forcing/ (solar-geophysical null result)