Crowdsourced aurora observations during geomagnetic superstorms: the Aurorasaurus open science database

1 Introduction

Aurorasaurus is an award-winning citizen science project that uses crowdsourced and citizen science data to create the first real-time, global aurora map (MacDonald et al., 2015). The platform and community comprise a long-standing gold standard for modern heliophysics participatory science (National Academies of Sciences and Medicine, 2020; National Academies of SciencesE. and Medicine, 2025), and are now collecting data in a second solar cycle. Significant earlier results include demonstrating the utility of the database (Kosar et al., 2018a; c), as well as contributions to the discovery and study of subauroral phenomenon STEVE (MacDonald et al., 2018; Semeter et al., 2020; Hunnekuhl and MacDonald, 2020). The project began in order to robustly collect data at scale during auroral events, with robust data quality control measures in place. The platform is limited to English, but widespread AI translation tools allow broader use. Citizen science projects generate a variety of assets, including science products, open data, methods of reciprocity, and embedded terms of use (Wiggins et al., 2018). To this end, Aurorasaurus utilizes an adapted Science Products Inventory for evaluation and tracking, as described in Edson et al. (2024).

Aurorasaurus conducts outreach and education across the globe, often via partnerships with local groups of enthusiasts (MacDonald et al., 2018). A global “superuser” cohort of citizen scientists, scientists, and students called the Aurorasaurus Ambassadors serve as collaborators. Evaluation results demonstrate successful connection with this highly motivated cohort, who represent the project and share Aurorasaurus and aurora science with their respective communities (Fischer, pers. comm. 2026). The Ambassadors themselves comprise a valuable resource for creating these opportunities, and the program leverages and reciprocates their local connections and enthusiasm by offering ways for them to build scientific knowledge and outreach skills through training and practice.

Aurorasaurus efforts have catalyzed participatory science growth in the field of heliophysics, and indirectly helped many successful projects of the Heliophysics Big Year (HBY) of October 2023–December 2024. The HBY, named for the “Big Year” concept in birding (Obmascik, 2011), was a NASA-led, global celebration of the Sun’s influence on Earth and the solar system. This initiative encouraged participants to engage with the Sun in as many ways as possible, recognizing the strong potential of a time at the nexus of citizen science, eclipses, and solar maximum. Aurorasaurus leadership initiated the concept, coining the term “Heliophysics Big Year”. Aurorasaurus collaborated on the HBY with NASA’s Science Activation (SciAct) teams, as well as broader HBY solar eclipse programming through a variety of education and outreach initiatives (Chambers et al., 2025). This paper will briefly describe such efforts, while the main topics are significant updates, an open science data release, and a discussion of the abundant Aurorasaurus data from two rare storms that occurred in 2024 during solar maximum.

The Sun’s most active period coincided with the HBY, yielding the remarkable superstorm of May 10–13, also known scientifically as the Gannon storm, and a second event on October 10–13 (Hayakawa et al., 2025; Fogg et al., 2026). A superstorm is typically defined as a rare geomagnetic storm with a minimum Dst index reaching −250 nT or lower (Tsurutani et al., 1992; Gonzalez et al., 2002)). The Gannon storm was driven by an unusual sequence of multiple coronal mass ejections, or CMEs (Hayakawa et al., 2025). The event was the largest of its kind since November 2003 (Grandin et al., 2024), and notably the first superstorm to be captured by widespread consumer digital cameras. The Gannon storm included coincident solar energetic particles and a rare ground level enhancement event, with accompanying extreme increases in D region and mesosphere ionization (Hayakawa et al., 2025; Papaioannou et al., 2025). The current solar maximum has also been the first to feature readily available smartphone software capable of capturing subvisual aurora, enabling the most extensive, global aurora reporting in the Aurorasaurus database to date.

Millions of people saw the Northern and Southern Lights during the Gannon storm. The importance of aurora citizen science was evident across the globe, reflected by many different reports and types of scientific efforts. Many of these works focused on gathering regional observations, e.g., 775 reports from Japan (Kataoka et al., 2024) and 45 reports from Mexico (Gonzalez-Esparza et al., 2024). In addition, heightened interest inspired many social media posts, reflecting that real-time online activity is known to scale by storm size (Case et al., 2015). After the Gannon storm, some users scraped tweets by hand for mapping (Wicklund, 2025). Such efforts are notable, but are not generally suitable for research use in the absence of permissions. Other researchers created surveys to gather reports post hoc and ask specific questions not covered by existing platforms like Aurorasaurus and Skywarden (Bruus et al., 2025). For example, 696 reports from over 30 countries were gathered via a survey by the Auroral Research Coordination: Towards Internationalised Citizen Science (ARCTICS) group (Grandin et al., 2024). Such efforts in general may fall under Human Subjects Research requiring Institutional Review Board review, as well as user information safeguards (Resnik, 2019; National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, 1979). In addition to information about location, color, and photographs, Grandin et al. (2024) asked users to self-report space weather effects, while Hayakawa et al. (2025) asked about naked eye sightings. Fogg et al. (2026) presented a comparative analysis of the May and October 2024 superstorms, examining their drivers, storm-strength models, geomagnetically induced currents (GICs), auroral activity, and magnetometer conditions over Ireland.

Global observations of aurora during rare superstorms are exceedingly scarce, so the behavior of aurora at low latitudes remains poorly understood (Ebihara et al., 2017; Hayakawa et al., 2023). Since the most intense storms are both rare and can be highly damaging, understanding the extent of auroral displays is an important proxy for storm strength, particularly for comparison to historical events (Knipp et al., 2021; Lockwood et al., 2025). In addition, rare types of aurora are themselves proxies for extreme manifestations of the magnetosphere under strong solar driving, e.g., a long-term mid-latitude study suggests that SAR arcs often mark the ionospheric footprint of plasmapause-ring current interaction (Mendillo et al., 2016). Participatory science data are well suited to powerfully inform these questions, as they are most abundant during these times. However, we note that citizen science data do contain population and geographic biases and are not appropriate for deriving absolute abundances or magnitudes. These caveats apply to many fields utilizing citizen science and are understood by practictioners.

This data release includes open-source software code to encourage inquiry among the public and enable broad collaborations. In parallel to space mission data best practice, a “Rules of the Road” document containing data usage guidance is also released. These are new guidelines for users that describe the terms of use for the data and an evolving understanding of how to best accommodate those terms. We note that not all citizen science data are suitable for release (de Sherbinin et al., 2021). We note that the Aurorasaurus terms of use protect the rights of photographers and the full photo dataset is not available in the public files, but photos are available to researchers who contact the team. In the nascent age of generative AI, it is important to protect photographers’ rights and data quality. Counterfeit images have not been a problem for Aurorasaurus to date, but the ease and sophistication of image generation is increasing as rapidly as trust in science and underlying infrastructures are being eroded. Having humans in the loop is essential (Fortson et al., 2026).

2 Overview of Aurorasaurus data

Aurorasaurus users submit reports at aurorasaurus.org. Each web report includes whether the observer saw or did not see the aurora; a timestamp of submission time and reported observation start and end times; geographic location; and a series of optional structured observational fields such as sky conditions, aurora colors, and aurora types. Users may also attach a photograph of their observation. Unlike earlier phases of the project that included content automatically extracted from Twitter social networks (Case et al., 2015; 2016a; Kosar et al., 2018a), this study focuses exclusively on reports submitted by users through the Aurorasaurus web platform. We note that prior to March 2023, Aurorasaurus scraped millions of reports robustly and automatically from Twitter (Case et al., 2016a), but the platform removed public API access as it transitioned to X. The loss of the Twitter reports, which included an effective process for verifying and validating them as real-time reports with an accurate location, was sizable as such ad hoc reports were on the same order of magnitude as the direct reports (Kosar et al., 2018a).

The Aurorasaurus project has been operational since September 2014. A previous technical report (Kosar et al., 2018a) documenting its early data coverage and 2015 and 2016 processing protocols highlighted the scientific utility of citizen science for real-time auroral boundary mapping. The present work builds on that foundation and includes a reproducible pipeline for data cleaning and scientific use, written in Python. This report includes the full project archive through July 2025, with particular emphasis on two recent extreme geomagnetic superstorm periods: the Gannon storm (May 10–13, 2024) and the October storm (October 10–13, 2024). These two events produced the highest volumes of reports in the project’s history, and serve as focal case studies for the data transformations and scientific analyses that follow.

Figures 1a,b show the breakdown of the raw reports submitted by users during two major geomagnetic storm events: the Gannon storm and the October storm. These pie charts illustrate three distinct categories of user behavior: (1) users who reported seeing the aurora and uploaded an image, (2) users who reported seeing the aurora but did not submit an image, and (3) users who submitted negative reports indicating that they did not see the aurora. During the Gannon storm, the platform received a total of 6,415 direct reports. To our knowledge, this is the largest number of sightings reported globally as participatory science for a single event. Out of those reports, 69.8 % are positive sightings and 30.2 % are negative sightings. As for the October storm, the platform received a total of 1,589 reports, of which 83.0 % were positive sightings. Notably, compared to the 2 years of data from 2015 to 2016 in Kosar et al. (2018a), a significantly higher percentage of reports (more than 50% compared with 20%) contained images, likely reflecting more widespread use of smartphones capable of directly capturing aurora. In the early years of the project it was common for users to submit “back of camera” shots taken by smartphone of the back of their DSLR camera’s photo of aurora, whereas now the smartphone itself can take the shot in the field. Additionally, the Aurorasaurus project had iOS and Android apps that provided significant quantities of reports, but software deprecation and maintenance issues limited use after summer 2018. A more mobile friendly software implementation has been released in 2026. An important issue in long-term citizen science infrastructure is that scientific data systems may outlive the rapidly evolving consumer software ecosystems they initially rely upon.

2.1 Description of Aurorasaurus data file content

The underlying structure, collection protocols, and file organization of the Aurorasaurus database have been extensively described in previous technical documentation (Kosar et al., 2018a). The hierarchical organization of raw data files remains unchanged and is not repeated in full here. The main data file under the name web_observations_2014–08-01_to_2025–08-02_cleaned.csv contains the cleaned reports from the project’s lifespan until August 2025, ordered by ascending “time_start” (the start time of the observation). Table 1 shows the data attributes in that data file and descriptions of each attribute. To make this dataset “cleaned” meaning ready for scientific use, the dataset underwent a structured, three-part processing pipeline: (1) data transformation, (2) data cleaning, and (3) advanced analysis and visualization. The entire workflow was implemented in Python using supporting scientific libraries: pandas (McKinney, 2010), matplotlib (Hunter, 2007), aacgmv2 (Shepherd, 2014), hapiclient (Weigel, 2021), cartopy (Met Office, 2013), seaborn (Met Office, 2013), and spacy (Honnibal et al., 2020)). This approach emphasizes Open Science principles such as open data, protocols, materials, and code (UNESCO, 2021; Wilkinson et al., 2016). The pipeline was modeled after the data preparation protocols described in Kosar et al. (2018a), and is designed to ingest raw comma-separated value format files and output a standardized and cleaned database in the same format. Relevant files, including documented Jupyter Notebooks that guide the user through each step, are hosted in the accompanying Zenodo repositories and linked Github directory (MacDonald et al., 2026; Kennedy et al., 2026).

TABLE 1. Data attributes in transformed and cleaned CSV Data Files.

Each report includes geospatial metadata derived from browser-based location services. As noted in Kosar et al. (2018a), geographic coordinates were used to calculate derived geomagnetic parameters, specifically magnetic latitude (MLAT), magnetic longitude (MLON), and magnetic local time (MLT), using the AACGM-v2, library. Duration and geomagnetic parameters are calculated in the cleaning process and are only included in the cleaned file.

The data transformation at the start of the pipeline transforms the raw Aurorasaurus backend database into a format consistent with the structure and field definitions of the data files in Kosar et al. (2018b). In this stage automated screening procedures and manual review identify test reports, spam, and personally identifiable information in text input fields. These are removed or redacted as appropriate. The second step is data cleaning, which ensures scientific quality and applies several filtering procedures to the raw dataset. The first filter removes reports that exceed 12 h in duration, which was determined to be the longest appropriate time for an overnight observation; longer observations likely indicate a data entry error. This is an addition to the Kosar et al. (2018b) process, which did not filter by duration until later in the scientific analysis. The second filter removes negative reports in which users selected that sky conditions were cloudy or bright due to light pollution or moonlight. This filter is applied because our primary scientific analyses rely on negative reports from observers with a clear view of the night sky, as obscured conditions cannot confirm or rule out the presence of an aurora. The optional third filter removes duplicates identified via an automated screening and manual resolution process. For the Gannon storm, we tested the third filter and described the process below to document results, but since the number of duplicates was very small the process was not conducted on the full dataset.

For the Gannon storm, we developed an automated protocol that flags observations with identical entries for a subset of the data attributes described in Table 1. The data attributes were applied to both positive and negative reports, and included observation start time, end time, country, state, and selected answers to the standardized form questions. Next, we manually reviewed the flagged reports, and classified each of them as duplicates (in pairs of which one entry was retained and one removed) or not duplicates. We introduced review protocols to aid in classification. Examining images was helpful as some identical entries had unique images and were thus not duplicates. Likewise, location coordinates proved useful, as duplicates were likely to be within ∼1 km of each other, though they would not appear identical due to each report having its own random offset for privacy. Differences in latitude and longitude between pairs of potential duplicates are shown in Figure 2b, illustrating that most duplicate reports were closer than an expected threshold estimate, while non-duplicates were further apart. Some edge cases illustrated limitations: for example, when a user submitted multiple photos as separate reports, they appeared largely below the (estimated) threshold “cut-off” line. In addition, clusters of reports within a small area may indicate population centers rather than duplicates, e.g., during the Gannon storm, reports were concentrated in England, where the aurora was visible. Similarly, non-sightings were clustered near major cities, possibly due to light pollution.

Figure 2a shows the results of the cleaning procedure and three filters described above. With additional manual classification of the “Flagged as duplicates” category 1.4 % of total reports were classified as duplicates, and retaining half of these, less than 1 % of the reports from the Gannon storm would be removed from the dataset. To test the reproducibility of our pipeline, we ran it on the 2015–2016 data that comprised the case study in Kosar et al. (2018a), and found comparable numbers for both raw and cleaned data. We note that erroneous data can still be included in the cleaned data, e.g., due to data entry errors, system errors, or the like. Such errors are observed to be very rare and may be further filtered as needed by the user.

The final stage of the pipeline involves scientific analysis of the fully cleaned Aurorasaurus dataset. This advanced analysis program builds upon the earlier cleaning and filtering stages, and produces a range of quantitative visualizations and statistical summaries designed to assess various characteristics of citizen science submitted reports. At this stage, for the analyses of interest here, we further filter the data down to reports lasting less than 3 hours as was also done by Case et al. (2016b) to allow for timely comparison to geophysical data.

2.2 Citizen scientist descriptions of auroral observations

The pipeline data cleaning procedure was applied to the Gannon and October storms. Figures 3a,b show the breakdown of reports in the final, cleaned dataset for the two superstorms as compared with raw numbers. After the data cleaning procedure for the Gannon storm, 5,109 of the 6,415 reports remained. The majority of these (58.2%) were positive reports accompanied by images, 20.0% were positive but submitted without a photo, and 21.8% were negative reports. For the October storm, after cleaning the 1,757 raw reports 1,623 remained, with 61.9% of cleaned reports containing images, 32.3% being positive reports without images, and only 5.9% being negative observations. We note that the number of reports during the October storm is artificially low due to a significant website outage.

The Aurorasaurus website experienced difficulty during both storms due to high traffic. During the Gannon storm, there was a 1-h outage before the peak of the main storm phase. During the October storm, the site experienced a partial, 4-h slowdown at a similar high reporting time in the main phase, followed by a more severe outage for nearly 2 days, then another approximately 2-day partial outage. Operating websites like Aurorasaurus during times of extreme traffic is difficult and costly despite tools for scaling servers and managing load. Sustained investment in experienced developers is necessary to manage “tech debt” over time and maintain operations and user services. The outage intervals for both storms are included in Figures 4a,b, which show the number of reports per hour and the SYM-H index. SYM-H is similar to the Dst index, which describes the geomagnetic disturbances at mid-latitudes in terms of the ring current disturbance, but with a faster time resolution (Iyemori, 1990; Imajo et al., 2022). Data are plotted as a function of observation start time, including data that were submitted later than the outages. (We note that most reports submitted after the fact were positive). Although less intense in terms of geomagnetic activity (reaching −390 nT in SYM-H), the October storm nonetheless generated high volumes of user participation and provides an important comparative case for assessing user behavior, sky conditions, and auroral visibility at mid-latitudes.

Figure 5 shows a side-by-side breakdown of self-reported qualitative form content across four parameters: auroral color, auroral activity level, location in the sky, and aurora type. All questions on the form are optional. Each segment of the horizontal bars represents the relative frequency of a type of response within the corresponding form question, including whether a question was skipped (a “non-report”). Reporting the auroral color and aurora type allows the user to select more than one input, while auroral activity level and location in the sky are single selections from a dropdown menu.

Reports submitted during the October storm included more detailed descriptions across all parameters compared with the May event (e.g., only 9 % gave no answer on color in October versus 25 % in May). Notably, the form response level was more complete than for the data reported in Kosar et al. (2018a). This shift likely reflected both the evolving familiarity of users with the platform and the Aurorasaurus team’s efforts to guide users through the report submission process. For the multi-select parameters, we note that both storms were very colorful and the vast majority of users selected multiple colors. The reported auroral activity level also had a broad distribution. Reports from October described the aurora as “Quiet, not moving much” more often than May submissions (43 % vs. 39 %) which was consistent with the October storm being less extreme overall. (We note also that even during large events, “quiet” features like SAR arcs can be widely visible and reported.) Sky location results show that the Gannon storm had a more significant fraction of observations covering the whole sky ( 20 % ), or located directly overhead ( 18 % ). Similarly, the proportion of the “weaker” observations in the northern horizon and 45 ° N categories comprised a larger fraction in October. The percentage of “Discrete Arcs” and “Multiple Types” of aurora was roughly the same for both storms. Finally, aurora types such as “Diffuse Glow” and “Pulsating Patches” were fewer in May. Examining the distribution of these parameters during a storm could be a valuable, new way to characterize its evolution in future work.

3 Discussion of research relevance, engagement, and potential uses

Recent outcomes of the Aurorasaurus project have been the scientific applications and impact described in Section 3.1, as well as the education, outreach, and engagement impacts described in Section 3.2. The scientific value of the crowdsourced auroral observations collected by Aurorasaurus has been demonstrated in several studies spanning multiple disciplines. Prior work by Case et al. (2015) showcased the effectiveness of integrating social media data, particularly Twitter (now X), to detect real-time auroral activity during geomagnetic storms. The 2015 St. Patrick’s Day storm (minimum Dst − 223 nT) reports were evaluated against the OVATION auroral view line forecast produced by NOAA’s Space Weather Prediction Center (SWPC), revealing that citizen science observations frequently extended equatorward of the predicted auroral boundaries (Case et al., 2016b). Further analysis by Kosar et al. (2018c) compared these citizen science observations with empirical OVATION Prime 2013 (OP-13) and Zhang-Paxton auroral models, demonstrating that the OP-13 boundaries were slightly more consistent with the observer-submitted data. These studies can be extended by testing with new models and the much larger storms of 2024, but that is beyond the scope of the current work.

Figures 6a,b show the global distribution of cleaned Aurorasaurus reports during the Gannon and October storms, respectively. Most reports originated from mid- and high-latitude regions in the Northern Hemisphere, particularly over North America and Europe, where the platform’s user base was most active (MacDonald et al., 2015). Like many citizen science datasets, Aurorasaurus observations reflect uneven geographic participation and therefore may contain spatial reporting biases that should be considered in analysis and interpretation. However, there was still an abundance of first-time reports from unusually low latitudes, especially during the Gannon storm. The Southern Hemisphere showed less coverage, as there are fewer land areas where the aurora may be visible.

Hayakawa et al. (2025) and Lockwood et al. (2025) discuss the importance of the lowest latitude reports as a measure of storm strength that can be compared to historical reports prior to geophysical data. Hayakawa et al. (2025) emphasizes “naked eye” reports as invaluable for this purpose. Lockwood et al. (2025) posits that some light-adapted human eyes in dark sky locations are single-photon sensitive and thus that modern camera-only observations can be directly compared to historical “naked eye” only observations.

Aurorasaurus low latitude reports do not have a “naked eye” designation. The value of naked eye reports for comparison to the historical database was underestimated, in part due to the lack of historical “great” storms in the prior decade of development. However such reports are still relevant, according to Hayakawa et al. (2025) and Knipp et al. (2021). Lockwood et al. (2025) suggests that the naked eye is as good as a camera and suggests seven subcategories that would be needed for tracking naked eye reports. Thus, even with the ambiguity of a lack of such designations, our reports have value.

The furthest equatorward reports deserve a high level of scrutiny and detailed study. Aurorasaurus helped to document the details of an extremely low latitude camera-only report from Oman near 18 ° magnetic latitude as discussed by Lockwood et al. (2025). Two astrophotographers were taking overnight night sky photography in a dark sky site in Oman with multiple cameras. Upon looking at their footage, they realized they might have captured something unusual with red emissions. They contacted a NASA enthusiast and English translator in Oman, who contacted the lead author. The lead author then confirmed the time, location, and astrometry of their observations were consistent and seemed to be red aurora at the same time as one of the peaks of activity well-documented across Europe (E. MacDonald, pers. comm. 2024). Reports from atypical locations are tricky to document and it is best to rely on primary sources, or as close to the original source as possible, to confirm details.

The modern age of rapid, plentiful, and more easily collected reports can open up more of a “time-domain” look at the lowest-latitude aurora on hourly or sub-hourly time frames. This creates two advantages over reliance on (often singular) accounts of extreme lowest latitude. One is that simultaneous observation at multiple latitudes (not just the lowest) can take altitude into account more robustly. Secondly, modern observations tend to include additional identifying information on form, color, and patterns. By carefully integrating a multitude of heterogeneous observations, widespread low-latitude phenomena like SAR arcs may be tracked and distinguished from traditional storm-time red arcs extending to high altitude. Due to the scarcity of the largest, rarest storms, whether there are repeatable patterns, timing, and types of low-latitude aurora is an open research question. Such detailed global reports have been enabled by advances in technology. In the mid-19th century, telegraphs and expanding railroad networks increased the need for synchronized clocks across large geographic regions. The resulting improvements in standardized timekeeping also enhanced the precision of distributed scientific observations, including widespread auroral reports and geomagnetic measurements (Boteler, 2006; Blaise, 2001). Having an extensive database of rare events with more precise timing (such as the database described here) could allow deeper understanding of the time domain of storm evolution.

Figures 6c,d show the distribution of cleaned Aurorasaurus reports submitted during the Gannon storm as a function of absolute magnetic latitude (MLAT) and magnetic local time (MLT), respectively. Magnetic coordinates are important because they account for the position of the Earth’s magnetic pole with respect to auroral behavior; the Earth rotates under the auroral oval, which is typically broadest around “magnetic” midnight. The distribution of Gannon storm data as a function of magnetic latitude demonstrated a broad range of reports spanning from less than 30°and more than 60°MLAT, with submission hours primarily between 18:00 and 04:00 MLT with a peak around 21:00. Compared with (Kosar et al., 2018a), which examined 2015–2016 near the maximum of the last solar cycle, we noted a shift towards lower peak (49.5°), minimum (21°), and maximum (63°) latitudes, reflecting a stronger storm. This was expected and consistent with the significantly smaller overall activity of the last solar maximum, which resulted in an overall higher-latitude auroral oval region.

3.1 Example scientific applications

Auroral color is an important clue to understanding the energies, altitudes, and processes of specific forms. Citizen science reports of color are an important aspect of the Aurorasaurus dataset and provide critical scientific information. Green (557.7 nm) is the dominant auroral color, and results from electrons from space impinging on atomic oxygen in the atmosphere. These green emissions peak at about 100 km altitude where atomic oxygen is densest. Red aurora typically occurs in three major types during storms: 1) blood-red “great red aurora” easily visible to the naked eye, 2) Stable Auroral Red (SAR) arcs, which can range from subvisual to visual with a pure red (630 nm) atomic oxygen emission, and 3) high-altitude atomic oxygen emissions, which can be bright and seen from further distances—including just over the horizon, which blocks the accompanying lower altitude green aurora from view (Robinson et al., 1985; Kozyra et al., 1997). SAR arcs tend to have extraordinarily long duration, persistent lower latitude locations in both hemispheres, and distinctive pure red color (Kozyra et al., 1997). Recently, an enhanced equatorial aurora has also been described at extremely low latitudes (He et al., 2020). A pink auroral color occurs primarily through more energetic electron bombardment, which penetrates deeper into the atmosphere (through the atomic oxygen layer into the molecular nitrogen layer below) and forms a fast-moving lower auroral border below 1) typical green arcs or 2) the patches typical of pulsating aurora, at higher and lower latitudes, respectively. Blue color in aurora occurs through resonant scattering of sunlight at high altitude with ionized nitrogen molecules (Nanjo and Shiokawa, 2024; Remick et al., 2001). All of these colors can also combine as light appears to blend, yielding magenta, purple, and orange. Magenta or purple can come from the combination of blue and red light, as described in Kataoka et al. (2024).

An observer’s understanding of the physical processes that drive aurora, as well as factors such as darkness, sky viewing quality, human perception, and camera settings, can influence the way they report colors. While red, pink, blue, magenta, purple, and even green auroral colors can each be produced in multiple ways, some phenomena have distinctive location, time, and duration features that can be used to aid in distinguishing between similar types. For example, STEVEs also have distinctive features and occur in the same subauroral, rapidly flowing trough region as SAR arcs (MacDonald et al., 2018). There are rare observations of SAR arcs that evolve into the narrower, more rapidly flowing STEVEs, but this evolution is not well understood (Martinis et al., 2022). STEVE also has some intermittent green rays that are subtly not the same color as the typical green aurora in a way that highlights the underlying process has key differences (Mende and Turner, 2019; Harding et al., 2020). Other ionospheric processes, like highly variable ion outflow, are much less visible (Peticolas et al., 2002; Ilie and Liemohn, 2016).

During superstorms, ground observers may be looking up at numerous forms that can appear to blend together. Complicating this further is the fact that different emissions have significantly different times for the photons of light to be emitted. For example, the red light associated with oxygen takes nearly 2 minutes, whereas the green light associated with oxygen takes less than a second, and the pink light associated with nitrogen molecules takes a fraction of a second ( 1 0 − 8 s ) (Rees and Luckey, 1974). Proton aurora also causes red emissions (Lummerzheim et al., 2001). Depending on the exposure length of a photo, forms with shorter emission times may be blurred. In addition, different camera and editing choices can be made to alter colors while automatic image processing by smartphone cameras, including white balance, noise reduction, HDR, and color enhancement, can alter the apparent colors. Self-reported colors may also be subject to human perception and memory biases. Raw, unedited photos are preferred for scientific usage. It is beyond the scope of this paper to sort out all of the different independent factors, however, we wish to point out the ways in which citizen scientist photos can be used to aid in these distinctions in combination with other information, such as that provided by the standard forms used by Aurorasaurus.

When the aurora extends to lower latitudes there are fewer scientific-grade, multicolor cameras to document it, so large storms have never been fully imaged (MacDonald et al., 2015). In addition, the largest superstorms are rare, so the intricacies of the drivers behind the unusual types and colors of aurora experienced during 2024 are not well understood. Viewers around the world observed long-lasting SAR arcs, time periods of great red aurora, and some over-the-horizon red aurora. Some locations saw multiple types at different times throughout the night, and some saw them at the same time but different parts of the sky. They also saw incredible magenta aurora, and more unusual aspects and combinations of these types. Researchers have investigated these with citizen science data (Grandin et al., 2024; Gonzalez-Esparza et al., 2024; Hayakawa et al., 2025; Nanjo and Shiokawa, 2024). The Aurorasaurus database contains plentiful photographic and qualitative information about these types, which can be filtered to help document which types were seen at which times. This can be combined with other ground- and satellite-based instruments to deepen our understanding and improve models.

A simple example is shown in Figure 7a, which presents the distribution of auroral color observations reported during the Gannon storm in the northern hemisphere, grouped by geomagnetic latitude (MLAT) in 0.5°bins. Similar to Grandin et al. (2024), the reports were separated by color to provide insight into the spatial extent and relative frequency of different auroral features described by observers. The figure plots every occurrence of each color: red, green, pink, and white were counted from the multi-select question about auroral color, while blue and purple were counted by scraping the “Additional comments” text box for case-insensitive, partial matches of both words. Color descriptions included magenta and, more commonly, purple; we considered these largely synonymous for the purpose of this study.

Green, pink, and red were most frequently reported during the May event, with peak reporting densities centered between 45° and 55°MLAT. Green aurora was most present, across the broadest latitude range. Pink and red were reported less frequently, but red was reported at lower latitudes. Blue and purple were also reported less frequently, but in sufficient quantities for trends to appear. In particular, we only plotted northern hemisphere data for these two storms, which occurred 6 months apart, at times of the year when there are significant differences in the amount of daylight. We hypothesized that viewers could see more blue aurora, as well as purple and pink, due to color mixing during the shorter nights of May. Conversely, we expected proportionally more red for the October storm with more darkness. This seasonality effect appeared consistent with both storms (Figures 7a,b), at least partially explaining the relatively more frequent reports of blue and purple during the Gannon storm. Southern hemisphere reports (not shown) were much fewer. However, especially during the Gannon storm, southern hemisphere participatory scientists seemed to report more red than the northern hemisphere, which was consistent with this hypothesis. We also note that since Aurorasaurus data (particularly for the May storm) were so numerous, they could be filtered by multiple criteria to attempt to aid in separating the many complicated types of red and pink aurora.

Despite being the two most significant geomagnetic storms of Solar Cycle 25, the May and October 2024 events have received limited direct comparative analysis to date. However, the availability of modern, well-calibrated instrumentation and extensive multi-domain observations enables opportunities for robust cross-comparison (Fogg et al., 2026). provide one of the few studies to explicitly examine both events, comparing their solar drivers, ground-based impacts, photographic records, and socio-historical context, with particular focus on geomagnetically induced current (GIC) signatures over Ireland. Their analysis identifies distinct solar drivers, showing that the October 2024 storm had a generally lower complexity and peak severity though reached comparable extremes for some indices and data. This interpretation is consistent with independent explanations summarized in the Aurorasaurus blog (K. Nykyri, pers. comm. 2024), which attribute the greater geoeffectiveness of the May storm to interacting CMEs and prolonged, variable southward IMF, producing Kelvin-Helmholtz waves in contrast to the more isolated and structurally simpler driver of the October event (Nykyri, 2024). Aurorasaurus operates as a real-time citizen science platform that actively responds to community observations and questions during ongoing space weather events—often on timescales far faster than formal scientific publication. While such communications are not peer-reviewed, they provide valuable contextual synthesis during active events, and the blog post additionally includes a global set of citizen-submitted auroral photographs, illustrating the widespread observational reach of both storms (Aurorasaurus, 2024).

3.2 Importance of education and outreach to Aurorasaurus and the Heliophysics Big Year

The Heliophysics Big Year encompassed not only solar eclipses, but solar maximum: the most active period in the Sun’s 11-year cycle. By doing so, it created a number of unique opportunities for education and outreach collaboration. For example, Aurorasaurus collaborated with the NASA Heliophysics Big Year team on a freely available lesson plan in which students created science communication “zines,” handmade publications with a long history in informal education (NASA Heliophysics, 2023). Combined with citizen science efforts, this activity served as a powerful heliophysics education tool, with opportunities for lasting impact in classroom zine libraries, museum programming, and more. Likewise, Aurorasaurus also collaborated with open source interactive data visualization software OpenSpace. The OpenSpace team collaborated with NASA’s Community Coordinated Modeling Center (CCMC), as well as NASA Science Activation at the American Museum of Natural History. The collaboration created an interactive data visualization showing the progression of the May 2024 aurora superstorm via Aurorasaurus reports, extending the reach and impact of the largest solar storm of the Heliophysics Big Year (Resnick et al., 2026).

Citizen scientists contribute their valuable time, unique knowledge, and persistent efforts to the field of heliophysics, even beyond regular aurora reporting. The Aurorasaurus Ambassador group of scientists, students, and enthusiasts brings a wealth of interdisciplinary skills to the study of aurora. Expert speakers are invited to give presentations to the Ambassadors that are then posted to the Aurorasaurus YouTube channel, generating an ever-growing, global resource library that contributes to increasing and sharing knowledge about aurora research and the process of doing science.

Aurorasaurus also hosts public programming throughout the year, providing multifaceted opportunities for the public to engage with the Heliophysics Big Year. Virtual journal club activities posted to YouTube work to increase citizen scientists’ understanding of the process of science and scientific research. Semiannual online Report-A-Thons engage aurora chasers, scientists, and enthusiasts from across the US and around the world to build community and generate significant volumes of citizen science data. These were especially successful in gathering backdated reports from the 2024 superstorms. The Aurorasaurus scientific team continues to advise citizen scientist co-researchers on their research questions, as well as giving and coauthoring numerous talks and presentations at conferences and other formal venues. Educational talks to schools and informal learning groups (many associated with NASA’s Science Activation program) enable non-scientists to engage directly with the subject matter and scientific community.

As a culminating experience of the Heliophysics Big Year, in January 2025 Aurorasaurus collaborated with the Astronomical Society of the Pacific (ASP) to host the first Eclipses to Aurora Winter Field School at the University of Alaska Fairbanks (NASA, 2025). This week-long opportunity included expert lectures and hands-on opportunities to introduce these undergraduate and graduate students to space physics graduate studies in topics such as electromagnetism, observations of aurora, data reduction, citizen science, and instrumentation. In addition, students observed the aurora in dark skies, experienced participatory science fieldwork at high latitudes, and prepared ways to increase impact by sharing what they learned and experienced with their communities across the US. Through this unique program, the lessons of the Heliophysics Big Year will have helped prepare a new generation of scientists and inspire the wonder of aurora science across the nation.

4 Conclusion

During the rarest, largest geomagnetic storms, the spatial extent and dynamics of aurora are poorly documented and understood. Space weather as a field is bolstered by the efforts of thousands of volunteer citizen scientists (National Academies of Sciences and Medicine, 2020; National Academies of Sciences and Medicine, 2025). We present an overview of Aurorasaurus data during the two superstorms of 2024. As expected from prior events of varying magnitude, the amount of interest in reporting aurora is proportional to the size of the storm. During these events, the largest ever recorded in our database over a decade, the platform proved a robust method of valuable data collection despite difficulties with server load during peak times. In addition, the full record of cleaned observations submitted directly to the website have been reviewed and released, allowing for longitudinal comparisons over a solar cycle, using a set of new, open Jupyter Notebooks for easier access to data visualization and exploration.

The Gannon and October superstorms constitute two excellent and different examples to test models of extreme superstorms, totaling nearly 8,000 observations, more than half of which included a photo. These storms were formed by different combinations of extreme drivers, highlighting the variety of compound threat mechanisms for extreme events. Aurora were reported during the Gannon storm from lower than 25°magnetic latitude. One example of high scientific interest was the variety of aurora types in multiple rare hues: red (including SAR arcs and Great Red Aurora), blue (indicating heavier ion outflow dominant in large storms), and combinations thereof. We also saw significant differences between the superstorms of 2024, consistent with their different seasons and the relative impact of more darkness on visible colors.

The sustained power of cultivating participatory scientists and accompanying volunteered geographic observations through education and outreach are demonstrated by the passionate stewardship and development displayed by an avid and active subset of Aurorasaurus observers. These worldwide superusers can also provide space weather resilience, offer excellent science communication, and serve as exemplars to other members of the growing heliophysics citizen science community. Scientists can expect the public to demonstrate keen interest, agility to be in the right place at the right time, and observational skills at a level to notice unusual phenomena. Gains in science literacy, understanding aurora, and authentic engagement with disciplinary content related to NASA science and engineering are user-centered design goals for the project as a whole that also benefit the wider heliophysics field and our technology-reliant society. At its best, this work empowers citizen scientists to know that they can contribute to active scientific study and learn more. Protocols for participatory science are strongest when they have been developed and beta tested to ensure feasibility before wider use. Two-way communication with participants reveals strengths and opportunities. When protocols are co-developed, citizen scientists are better able to take “gold standard” observations in a more quantitative way, e.g., with fixed settings, clock checks, color calibration, and a fisheye lens when possible (Herlingshaw et al., 2024). They can also provide guidance with regard to fieldwork safety and feasibility. The Heliophysics Big Year was able to capitalize on lessons learned from many older citizen science projects as new ones were built for eclipse opportunities. Bringing together space weather enthusiasts around the globe promises to drive growth in heliophysics and open science discoveries. Another goal was to sustain interest in this science beyond the eclipse, e.g., through the equally stunning and widespread opportunities for aurora, driving home the power of stewardship for heliophysics as a field and a brand.

As technology advances, understanding space weather becomes more critical than ever. Participatory science initiatives like Aurorasaurus empower the public to be part of groundbreaking research, fueling discovery for humanity’s future. A goal of Aurorasaurus is to move the dial on evergreen challenges of communicating aurora in the public sphere, an effort has now rippled out and multiplied throughout the Heliophysics Big Year. Through the enthusiastic participation demonstrated in 2024 the door has been opened to a new generation asking even deeper questions. Participatory science projects like Aurorasaurus connect research toward practical solutions for our technological world with wonder at the night sky, nurturing the links between humanity and nature. This feedback loop of greater understanding leads to better data, which leads to greater understanding.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Help troubleshooting the bibliography and Python code, incidental use for minor editing and research purposes.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Gareth Perry, New Jersey Institute of Technology, United States

Reviewed by: Alexei V. Dmitriev, Lomonosov Moscow State University, Russia

Alexandra Fogg, Dublin Institute for Advanced Studies (DIAS), Ireland