| 16 Dec 2025
Dune aurora: Statistical survey from a citizen science database
Abstract. Auroral forms can provide information not only on the state of near-Earth space but also on conditions in the lower-thermosphere–ionosphere. The so-called dune aurora, consisting of brighter stripes forming a wave-like pattern in the dim, diffuse green aurora, has been hypothesised as being an optical signature revealing the presence of large-scale atmospheric waves above or near the mesopause. However, only a few dune aurora events have been studied to date, leaving many open questions regarding the nature of this phenomenon. We carry out the first statistical analysis of dune aurora events by collecting citizen science observations of the dunes since 2000 using the Skywarden (https://taivaanvahti.fi) database of observations. From a total of 289 dune aurora observations made during 56 different events by citizen scientists from Northern Europe, North America, Australia, and New Zealand, we investigate the distribution of dune events as a function of location, month, local time, solar wind and interplanetary magnetic field (IMF) conditions, and geomagnetic activity. We compare those distributions to that of all the aurora observations reported in Skywarden since 2000. We also estimate the duration of dune events based on the available observations, and we investigate a possible relationship between dune aurora and equivalent current patterns derived from ground-based magnetometer measurements. We find that the vast majority of dune observations take place near the equatorward boundary of the auroral oval, in the dusk sector earlier than the peak in all auroral report distribution, and in association with strong (in most cases eastward but occasionally westward) auroral electrojet signatures. The dune observations are often associated with elevated solar wind density and IMF magnitude, and the IMF By component may play a role in their formation. Finally, their monthly distribution peaks in March and October, which could be the result of a combination of geomagnetic, atmospheric, darkness, and cloudiness conditions needed for them to form.
Summary
This manuscript uses the Skywarden citizen-science database to study the occurrence characteristics of dune aurora, with the aim of constraining formation mechanisms. The scientific motivation is clear. The Skywarden community effort itself is exceptional: sustained and careful documentation by many aurora photographers has created a valuable record of rare and subtle auroral forms, and the authors’ attempt to integrate this resource into research is timely.
My main concern is about what can be supported as “statistics” with this dataset. The manuscript acknowledges that biases exist, and I agree the authors are aware of that. However, in its current form the paper still presents “occurrence dependences” (MLT/season/activity/solar wind context) in a way that can easily be read as statistical properties of the phenomenon. Given the dataset structure, these quantities are not well-defined, and there is a risk that such results will later be cited as established occurrence trends even though they mainly reflect reporting and detectability effects. Below I outline the specific issues.
Major concerns
1) Occurrence inference without observational effort (a denominator)
To infer occurrence rates or systematic dependences, it is necessary to define the observational effort: when/where the phenomenon could have been detected, and under what observational effort. Skywarden is excellent as a collection of reported cases, but in general it does not include the kinds of information that are easily available in scientist-run observations, such as:
Without an effort model, or at least a defensible proxy, the results describe the distribution of reports rather than the occurrence properties of the phenomenon.
2) Reporting and detectability biases are not bounded; normalization does not remove them
Dunes are low-contrast structures, and detectability depends strongly on observer behavior, equipment/settings, and retrospective inspection. Reporting probability is also likely to vary with local time, geomagnetic activity, and community attention. For example, once brighter and more dynamic aurora appear (e.g., breakup arcs), observers may change pointing, switch targets, or stop observing because they feel they have already captured what they wanted, which can imprint an apparent MLT preference even if the underlying phenomenon does not.
I understand the authors’ approach of normalizing dune reports by the total number of aurora reports, with the goal of reducing sampling/reporting bias. However, dividing one biased distribution by another biased distribution does not, in general, remove bias. Many of the relevant biases (observer activity patterns, stopping rules, pointing choices, visibility/detectability, and social amplification) affect both the numerator and the denominator, and not necessarily in the same way. Even if some bias components are shared between the two, cancellation in a ratio only occurs under restrictive assumptions—essentially when the same multiplicative bias factor applies to both quantities in the same way and is independent of MLT, activity level, and other variables of interest. Without a way to quantify or bound these effects, the reported MLT/season/activity dependences cannot be interpreted as intrinsic properties of dune aurora.
3) Event non-independence and potential double counting
Multiple submissions may refer to the same physical event, especially after community sharing triggers retrospective searches. The manuscript seems to be aware of this and attempts some handling, but there is no rigorous event-level consolidation. Without event clustering and verification (e.g., geometric checks using star fields), the effective sample size and the shapes of distributions remain uncertain. In that situation, distributions can be dominated by repeated sampling of a small number of events, which weakens any statistical interpretation.
4) Ambiguity in timing and duration definitions
Times are derived from mixed sources (EXIF and narrative descriptions). Remaining timing uncertainties are not quantified, nor is their impact on MLT assignment and on the OMNI/SME/SMR averaging windows assessed. Duration estimates are additionally affected by different cadence/continuity and variable “presence” criteria across observers; this is not a standard measurement-error setting. As a result, duration statistics provide limited physical constraint as currently presented.
5) Interpretation of solar-wind/geomagnetic parameter distributions
Comparisons between dune reports and all aurora reports are presented as evidence for preferred parameter ranges or structured distributions. However, given (i) unknown observational effort, (ii) reporting bias that varies with activity and time, and (iii) non-independence/double counting, fine-scale distribution features (e.g., dips near specific IMF values or apparent multimodality) are not compelling as physical selectivity. More importantly, this is a structural limitation of using Skywarden alone for “occurrence statistics”: without effort constraints and event independence, such distributions are hard to interpret physically and should not be presented as robust statistical properties.
6) Latitude: observer location vs. dune location
The geomagnetic latitude used in the analysis is that of the observer rather than the dune emission region. If the goal is to understand dune behavior, the relevant quantity is the location of the phenomenon, not the location of the photographer. This conflation weakens latitude-related conclusions and also reduces the ability to de-duplicate events across sites. Since stars are visible in many images, approximate mapping/geolocation (with altitude assumptions) seems feasible in principle and could strengthen both dune-location inference and event consolidation.
7) SECS/equivalent current section: selection criteria and leverage
The criteria for selecting events for the SECS analysis are phrased qualitatively (e.g., “several clear observations”, “a very large number of observations across a broad area”), which makes the selection hard to reproduce. I do think the point that dunes are not restricted to only westward electrojet context (i.e., they can also occur with an eastward electrojet) could be an interesting result. If the authors want to make that a key message, it may be stronger to present it through a smaller number of well-defined case studies with clear selection rules and deeper analysis, rather than embedding it within a broad statistical survey.
8) Discussion scope
The Discussion covers many mechanisms but would benefit from clearer prioritization tied to what the present dataset can and cannot constrain under these limitations.
Minor comments
Overall assessment
Skywarden is extremely valuable for demonstrating the existence and diversity of dune aurora and for collecting candidate events. However, unless observational effort can be constrained and events can be consolidated at the event level, it is difficult to present the manuscript’s “occurrence dependences” as statistical properties, and it is risky to leave them in the literature as a basis for physical interpretation. A clearer path would be either (i) to reframe the paper as a descriptive catalogue / case-based study and stop making claims about statistical occurrence characteristics, or (ii) to analyze a scientist-run dataset with well-defined observational effort if the goal is to find the statistical behavior of the phenomenon.