---
source_pdf: Unexplained_Bright__Dim_Intensity_Asymmetries_in_SOHO_and_UVCS.pdf
title: "Draft version September 30, 2025"
site: https://densityfielddynamics.com/
author: Gary Alcock
framework: Density Field Dynamics (DFD)
format_note: "Markdown extracted from the source PDF for clean AI ingestion. No editorial changes; mathematical typography in the PDF is authoritative."
---

Draft version September 30, 2025
Typeset using LATEX twocolumn style in AASTeX631

Unexplained Bright–Dim Intensity Asymmetries in SOHO/UVCS Lyman-α Data
Gary Alcock1
1 Los Angeles, California, USA

ABSTRACT
We present a permutation test analysis of 334 daily sequences of SOHO/UVCS Lyman-α spectra
(2007–2009). By splitting frames into bright and dim subsets and comparing their median intensity
and wavelength, we test whether observed differences exceed those expected from random sampling.
We find that 163 of 321 day–radial bins (51%) exhibit statistically significant bright–dim intensity
contrasts at false discovery rate (FDR) 5%. The effect is modest (Cohen’s d ≈ 0.24) but robust to
permutation. Velocity differences are not significant (d ≈ −0.03). The origin of this asymmetry is
unknown: it may reflect solar structures, instrumental systematics, or unexplored physical processes.
We publish these results as an open anomaly to encourage community investigation.

Keywords: solar corona — spectroscopy — statistical methods — ultraviolet astronomy
1. INTRODUCTION

The solar corona exhibits complex dynamics observable through ultraviolet spectroscopy. The SOHO Ultraviolet Coronagraph Spectrometer (UVCS; Kohl et al.
1995, 1997; Raymond et al. 1997) provides long time–
series of Lyman-α observations. Instrumental stability
has been studied in depth (Gardner et al. 2002; Strachan et al. 2002; Guhathakurta et al. 1999; Kohl et al.
2006), but statistical anomalies may persist in archival
data. Here we document bright–dim intensity asymmetries in UVCS daily sequences using a nonparametric,
permutation-based approach with multiple-testing control.
2. DATA AND METHODS

We analyzed 150,685 frames from 895 UVCS exposures (2007–2009). Frames were binned by day and
projected radius (Rbin = 0.0 corresponds to the innermost detector bin, ≈ 1.5–1.6 R⊙ ). For each (day,Rbin )
with ≥ 2 frames, we sorted frames by total line intensity
and split at the median, with the upper half designated
“bright” and lower half “dim”. This median-split definition replaces earlier descriptions of “random partitioning” to ensure reproducibility.
For each group we computed:

Permutation tests (20,000 replicates per group; Ernst
2004; Good 2013) generated null distributions; two-sided
p-values were corrected by Benjamini–Hochberg FDR
(Benjamini & Hochberg 1995). Effect sizes were quantified with Cohen’s d (Cohen 1988). Instrumental filters
used in exploratory work (−0.9 < ∆I < 20, |∆v| < 120
km s−1 ) were not applied in the final analysis; results
reported here use the full data without post-hoc rejection.
Data quality notes: Daily frame counts ranged from
2 to 150 (median ∼ 20). We verified that all exposures
passed UVCS quality flags as archived by the SOHO
Science Archive (SSA). Known pointing offsets and detector artifacts are logged in Gardner et al. (2002), but
no anomalies specific to the intervals studied (2007–2009
solar minimum) were documented.
2.1. Pre-registered flagged-day definition
To avoid circularity, we pre-registered a binary
“flagged” label (hereafter “flagged (significant) days”)
for external-validation analyses: a day is flagged if the
permutation test for intensity contrast at Rbin = 0.0
yields a two-sided p < 0.05 (before any comparison to
external solar activity catalogs). This label is then used
once, as an independent predictor in the CME coincidence analysis below.

1. Intensity contrast ∆I = (Ibright /Idim ) − 1,
2. Wavelength shift ∆λ = λbright − λdim ,
3. Doppler velocity ∆v = c ∆λ/λ0 , with λ0 =
1215.67 Å.

2.2. CME time+angle coincidence scoring
We assessed external validity by cross-matching UVCS
observing windows with LASCO CMEs (Brueckner et
al. 1995; Yashiro et al. 2004). For each UVCS day

2

Alcock

we constructed a binary indicator aligned any that
is 1 if a cataloged CME occurred within a temporal
padding window (pad ∈ {0, 30, 60, 120} minutes) of the
UVCS observing interval and within an angular tolerance (tol ∈ {0◦ , 5◦ , 10◦ , 15◦ , 20◦ , 30◦ }) of the UVCS slit
position angle. We then computed the aligned any
rate separately for (i) all days and (ii) the flagged
subset (Sec. 2.1), and summarized the rate difference
(flagged minus all) across the (pad, tol) grid. Uncertainty and enrichment significance were gauged with
Fisher’s exact test on the 2 × 2 table of hits/non-hits
in the two groups. The per-cell results are packaged in
cme timewindow significance grid.csv (Data Availability).
Note on GOES flare coincidence. We attempted an
analogous analysis using daily SWPC text files for
GOES flares. For 2007–2009, the specific daily endpoints we probed returned unavailable (HTTP 404) or
empty files, yielding no usable lines. Because this epoch
lies in the deep solar minimum and the archive access
was incomplete for the daily text products, we treat flare
coincidence as non-informative here and rely on CMEs
as the independent coronal activity proxy.
3. RESULTS

Out of 321 testable day–radius groups, 163 passed the
5% FDR threshold for intensity contrast. Figure 2 shows
the permutation p-value spectrum, indicating clear departure from the uniform null. Figure 3 displays intensity contrast versus velocity shift; significant cases cluster at nonzero contrast but near-zero ∆v. Figure 4 compares observed and null distributions. Figure 5 shows
the temporal distribution of significant cases; a compact
year-by-month view appears in Table 2.
Why only Rbin = 0.0? The UVCS campaign density
during 2007–2009 yields robust sample sizes at the innermost bin only. Higher Rbin locations did not provide
sufficient (day,bin) groups with ≥ 2 frames to support
a reliable permutation test and FDR control; counts
are summarized in Appendix A. We therefore report
Rbin = 0.0 as the adequately powered subset and document the absence of higher-radius detections for completeness.
Overall effect sizes: contrast d = 0.24 (small–
medium), velocity d = −0.03 (null). A simple power
analysis indicates that N ≳ 10 frames per day are required for robust detection at d ≈ 0.24 with α = 0.05
(Faul et al. 2007). A frame-count histogram is provided
in Figure 6 to visualize sampling depth across days.
3.1. External validity: CME coincidence

Figure 1. External validation via LASCO CME coincidence. Heatmap shows absolute enrichment in CME
time+angle alignment rates (flagged minus all) as a function
of time pad (minutes) and angular tolerance (degrees). Cell
annotations display the enrichment values (fractional units).
A color bar indicates the magnitude and sign. Per-cell twosided Fisher p-values and permutation p-values are supplied
in the CSV cme timewindow significance grid.csv.

Flagged (significant) days (Sec. 2.1) exhibit consistently higher CME time+angle coincidence than the
full set across the (pad, tol) grid. Figure 1 summarizes
the enrichment as absolute rate difference (flagged minus all); enrichment is positive in all displayed cells,
with typical magnitudes ∼0.10 for pad= 0 min and
∼0.15–0.25 for pads of 30–120 min. As a representative configuration (pad= 60 min, tol= 10◦ ), the baseline
CME-coincidence rate for all days is 0.606 (206/340),
the flagged-day rate is 0.786 (11/14), yielding an absolute enrichment of +0.180; Fisher’s exact two-sided
p = 0.263 and a label-shuffle permutation p = 0.265 for
the same 2 × 2 margins. Across the full 4 × 6 pad×tol
grid, all 24 cells show positive enrichment; a binomial
sign test against random ± signs gives p ≈ (1/2)24 ≃
6 × 10−8 (p < 10−6 ). The mean and median enrichments across the 24 cells are both in the +0.2–+0.3
range (see CSV). The number of flagged days is modest
(n = 14 at raw p < 0.05), which limits per-cell power,
so we emphasize the grid-level consistency rather than
any single-cell p-value. The high baseline coincidence
(e.g., ∼60% at pad= 60 min, tol= 10◦ ) likely reflects a
combination of LASCO catalog completeness and generous time/angle windows; the additional +18 percentage
points in flagged days indicates a substantial relative lift
rather than a floor effect, suggesting their asymmetries
are systematically linked to CME occurrence rather than
random chance.

3

UVCS Bright–Dim Asymmetries
Table 1. Strong Effect Sizes (Some Not Significant at FDR 5%)
Day

Rbin

26-MAR-2008
16-JUN-2009
27-MAR-2008
16-JUN-2008

0.0
0.0
0.0
0.0

∆v (km s−1 )

∆I
19.8
11.9
7.3
5.8

4.3
24.2
-9.6
22.5

p

qFDR
−4

< 10
< 10−4
< 10−4
< 10−4

< 10−3
3 × 10−4
1.6 × 10−2
3.1 × 10−1

Note—Rows illustrate cases with strong contrasts. Not all are significant at the 5% FDR level (e.g., 16-JUN-2008). We include them
for effect-size context; only q ≤ 0.05 entries are formally significant.

Table 2. Monthly counts of CLEAN FDR-significant bright–dim asymmetries
Year

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Total

2007
2008
2009
Total

9
5
···
14

···
8
···
8

···
10
4
14

2
11
6
19

6
6
5
17

6
10
2
18

1
5
2
8

···
···
5
5

11
13
7
31

2
1
4
7

9
···
···
9

13
···
···
13

59
69
35
163

Figure 2. Histogram of permutation p-values for contrast
differences. Departures from the uniform null indicate significant asymmetries.

Figure 3. Observed intensity contrast vs. Doppler shift
(clean subset). Significant cases cluster at nonzero contrast
while ∆v ≃ 0.

2002; Kohl et al. 2006); (3) unmodeled physical processes. We emphasize this is not definitive evidence for
new physics.
4. DISCUSSION

The bright–dim contrast signal is statistically robust,
yet its physical origin is uncertain. Possible explanations
include: (1) solar structures (e.g., streamers, CME remnants; Ofman et al. 1997; Raymond et al. 1997; Strachan
et al. 2002); (2) instrumental systematics (slit illumination, vignetting, detector response drift; Gardner et al.

4.1. Physical link between CMEs and bright–dim
asymmetries
CMEs introduce strong, evolving density and topology
changes in the low corona that can modulate Lymanα brightness via resonant scattering and line-of-sight
geometry. Density enhancements and altered illumi-

4

Alcock

Figure 4. Observed vs. null distributions of contrast and
Doppler velocity. Observed contrasts are systematically
higher.

to accommodate (i) CME onset/identification timing in
catalogs and (ii) propagation to 1.5–1.6 R⊙ . Typical
CME speeds of ∼300–1200 km s−1 imply transit times
of order 15–60 minutes across several tenths of R⊙ in the
low corona, though acceleration phases and projection
effects broaden this range (Zhang & Dere 2006; Temmer 2021). The positive enrichment at modest pads
suggests the asymmetry co-occurs with, or shortly precedes/follows, cataloged CME activity; the effect does
not require precise timing at the minute level to be detectable in our daily aggregates.
4.3. Negative controls and selection considerations

Figure 5. Timeline of significant bright–dim asymmetries
(CLEAN FDR 5%). Clustering is evident during specific
months.

To probe selection effects, we performed label-shuffle
controls that preserve the number of flagged days and
their epoch distribution (month-level counts) and recomputed the pad×tol grid. The resulting enrichment
centered near zero (as expected under the null), supporting that the observed CME coincidence is not an artifact
of epoch-dependent catalog completeness. We also note
that the CME method compares flagged days directly
to the full set drawn from the same calendar interval
(2007–2009), mitigating long-term activity-cycle confounds. Further geometric controls (e.g., cross-checks
of LASCO C2 position angles relative to the UVCS slit)
are natural extensions for future work.
4.4. Testable Hypotheses and Diagnostics
Specific hypotheses worth testing include:
• Instrumental: correlation with detector temperature telemetry; slit position and spacecraft roll
angle.
• Solar: coincidence with streamer locations from
LASCO C2 coronagraph images; periods of elevated solar wind speed (e.g., ACE/SWICS).

Figure 6. Distribution of frame counts per day (all Rbin ).
Vertical line marks median (∼20).

nation anisotropy can bias median-split intensity partitions without requiring large net Doppler shifts, consistent with the near-zero ∆v we observe. CME-driven
changes to the ambient streamer belt and sheath structures can therefore produce bright–dim intensity asymmetries even when velocity centroids remain stable (e.g.,
Yashiro et al. 2004; Strachan et al. 2002).
4.2. Timing and causality
Our coincidence pads (0–120 min) are symmetric
around each UVCS observing window and are intended

• Orbital/temporal: possible 6-month SOHO orbital periodicity; possible 27-day solar rotation cycle.
• Effect-size linkage (prediction): test whether
larger ∆I correlates with closer CME timing
and/or smaller angle offsets; a positive association
would further support a causal connection.
Future work should extend analysis across all radial bins
as data permit, and perform joint correlations with solar activity indices (sunspot number, F10.7 flux). Engagement with the UVCS instrument team is essential,
as they may immediately recognize these signatures as
known systematics or observing-program effects.

UVCS Bright–Dim Asymmetries

5

Table 3. Multi-radius summary
(CLEAN FDR 5%)
Rbin

Ngroups

Nsig

Frac. sig

0.0

334

163

0.488

5. CONCLUSIONS

We report unexplained bright–dim asymmetries in
UVCS Lyman-α data. The signal is statistically significant but modest, and lacks corresponding velocity
shifts. We release these results to the community as an
anomaly for further investigation.
DATA AVAILABILITY
The SOHO/UVCS data analyzed here are publicly
available from the SOHO Science Archive (SSA). Derived tables used in the figures and appendices are
included in the supplementary package (uvcs perm
results.csv, uvcs perm clean significant.csv, uvcs perm
clean top.csv, and summary CSVs in paper/data/). All
figure assets referenced in this manuscript (uvcs pvals
hist.png, uvcs contrast vs dv.png, uvcs obs vs null.png,
uvcs significant timeline.png, uvcs significant monthly.
png) are provided.

For the external-validation analysis, we include cme
alignment grid.png (Fig. 1) and the underlying grid table cme timewindow significance grid.csv with columns
{pad min, tol deg, k all, n all, rate all, k flag, n flag,
rate flag, enrichment, p fisher two sided, p perm two
sided}. Daily GOES flare text archives were unavailable
for this epoch under the queried endpoints, so flare coincidence products are not used. The negative-control figure cme label shuffle null.png (Fig. 8) is also provided.
ACKNOWLEDGMENTS
I thank the SOHO/UVCS instrument team (J. L.
Kohl, L. Strachan, J. C. Raymond, and collaborators)
for making the UVCS data publicly available and for
their foundational instrument papers. I also thank colleagues who provided methodological feedback during
the development of this analysis. SOHO is a project of
international cooperation between ESA and NASA.

APPENDIX

A. MULTI-RADIUS ANALYSIS

Although the analysis emphasized Rbin = 0.0, we queried additional projected radii. The number of day–radial
groups and fraction significant at FDR 5% are summarized below; only the innermost bin had sufficient groups to
support robust inference in 2007–2009.

B. TEMPORAL CORRELATIONS

We compared the timeline of significant asymmetries (Figure 5) to monthly counts (Table 2) to visualize clustering; a
fuller treatment against external indices (sunspot number, F10.7, SOHO orbital phase) is left to follow-up. Preliminary
inspection shows weak clustering during mid-2008, not obviously aligned with sunspot activity.

C. NEGATIVE-CONTROL SHUFFLE TEST FOR CME ENRICHMENT

We visualize the label-shuffle null for the CME enrichment grid as a histogram of mean enrichments across the 24
cells (20,000 shuffles).

6

Alcock

Figure 7. Monthly distribution of CLEAN FDR-significant bright–dim asymmetries.

Figure 8. Label-shuffle negative control for the CME enrichment grid. We randomly permute the “flagged” labels (preserving
count and month-level distribution), recompute the pad×tol enrichment in each of the 24 cells, and plot the resulting distribution
of cell-wise enrichments (gray histogram; 20,000 shuffles). The distribution is centered near zero with a narrow spread, as
expected under the null. Vertical solid (dashed) lines mark the observed mean (median) enrichment from the real labels; both
lie well outside the shuffle spread, complementing the sign test in Sec. 3.1.

UVCS Bright–Dim Asymmetries

7

REFERENCES
Benjamini, Y., & Hochberg, Y. 1995, J. R. Stat. Soc. Ser.
B, 57, 289
Brueckner, G. E., et al. 1995, Sol. Phys., 162, 357
Cohen, J. 1988, Statistical Power Analysis for the
Behavioral Sciences, 2nd ed. (Routledge)
Ernst, M. D. 2004, J. Stat. Softw., 8, 1
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. 2007,
Behav. Res. Methods, 39, 175
Gardner, L. D., et al. 2002, ApJS, 138, 399
Good, P. 2013, Permutation Tests: A Practical Guide to
Resampling Methods (Springer)
Guhathakurta, M., et al. 1999, J. Geophys. Res., 104, 9801

Kohl, J. L., et al. 1995, Sol. Phys., 162, 313
Kohl, J. L., et al. 1997, Sol. Phys., 175, 613
Kohl, J. L., Noci, G., Cranmer, S. R., & Raymond, J. C.
2006, A&ARv, 13, 31
Ofman, L., et al. 1997, ApJ, 491, L111
Raymond, J. C., et al. 1997, Sol. Phys., 175, 645
Strachan, L., Suleiman, R., Panasyuk, A. V., Biesecker, D.
A., & Kohl, J. L. 2002, ApJ, 571, 1008
Temmer, M. 2021, Living Rev. Sol. Phys., 18, 4
Yashiro, S., Gopalswamy, N., Michalek, G., et al. 2004, J.
Geophys. Res., 109, A07105
Zhang, J., & Dere, K. P. 2006, ApJ, 649, 1100