As components of landscapes, geoheritage sites often hold symbolic and spiritual meaning, reflected in religious traditions that attribute sacred status to features such as rocks and caves (Kiernan, 2015). Geological processes and events are embedded in cultural memory through oral traditions and place names, which record volcanic eruptions, earthquakes, and tsunamis and, in some cases, contribute to long-term risk awareness and disaster preparedness (Ludwin et al., 2007; King and Goff, 2010; Fepuleai et al., 2017; Sousa et al., 2010; Planas-Batlle et al., 2023; Fontanella Pisa, 2024; Isoda et al., 2019). In addition, geoheritage sites can provide essential contextual information for archaeological research by reconstructing palaeogeographic settings and documenting past human–environment interactions (Panizza and Piacente, 2009; Melelli et al., 2016; Coratza and Hobléa, 2018; Fairchild and McMillan, 2007). Geological materials and landforms have also been widely used in the construction of settlements, infrastructure, and cities, forming the basis of urban geoheritage and offering opportunities to connect geosciences with everyday human experience (Boukhchim et al., 2018; Coratza et al., 2016; Kubalíková et al., 2020; Pica et al., 2018). Together, these cultural dimensions highlight the need for holistic approaches to geoheritage conservation and communication that integrate scientific, cultural, and societal perspectives.

Quaternary geoheritage sites encompass geological/geomorphological sites whose primary heritage values originate in the Quaternary period (Sayama, 2024). This concept represents a broader interpretation of geomorphosites (sensu Reynard et al., 2009), which defines aesthetic characteristics as one key (but not sole) attributes. In general, Quaternary geoheritage sites demonstrate three main characteristics (Reynard, 2009; Sayama, 2024). Firstly, many Quaternary geoheritage sites are dynamic, as they are strongly influenced by ongoing earth surface processes. Secondly, these sites can exist at various scales, ranging from those covering entire landscapes, such as alluvial plains, to smaller individual features such as single rocks or crystals. Finally, many Quaternary geoheritage sites exhibit strong connections with human culture, encompassing archaeological, historical, architectural, and other dimensions. These sites were formed, are forming, or will be formed, concurrently with human history. With this overlap in natural and human history, Quaternary geoheritage sites can provide geoarchaeological records of the human-environment interactions over time and reveal human impacts on this nature-culture relationship. Moreover, Quaternary geoheritage sites have inspired creative ingenuity in art and cultural traditions, as exemplified by the religious and artistic significance of Mount Fuji, Japan (Oguchi and Oguchi, 2010; Chakraborty and Jones, 2018), or the importance of Holocene meteorite impact craters in Australia for Aboriginal Dreamtime stories, some with an oral history dating back more than 4500 years (Hamacher and Goldsmith, 2013). Given these distinctions, many geoheritage studies (Moradi et al., 2021; Pereira and Pereira, 2010; Pereira et al., 2007; Erhartič, 2010) have treated Quaternary sites (or geomorphosites) separately from older geoheritage sites, particularly in site evaluations (Santos et al., 2020; Mucivuna et al., 2022).

The lack of appreciation of geosciences and geoheritage has been documented from various perspectives. For example, on social media, the hashtag “geology” has been used approximately 80 % less frequently than “physics”, “biology”, or “chemistry” (Zawacki et al., 2022). Even when the media features geosciences, Stewart and Nield (2013) found that topics broadcast on British television were usually limited to a few popular domains, including palaeontology, volcanology, and seismology, with a range of featured Quaternary topics mainly related to geohazards or archaeology. These findings highlight room for improvement in the public recognition of geosciences, including Quaternary science.

Effective communication strategies have garnered attention and recognition as potential solutions to this apparent “detachment” (Stewart and Nield, 2013) from geosciences. As described by Illingworth et al. (2018) and Rodrigues et al. (2023), geoscience communication is still at its infancy as an academic field. The non-specialists' difficulty in understanding deep time (Bowring, 2014; Warmold, 2017; Trend, 2001), unfamiliarity with technical jargon (Ren et al., 2013; Kortz et al., 2017), and the global lack of geosciences education in schools (Melendez et al., 2007; Reis et al., 2014; Subedi et al., 2020), have been identified as primary hurdles to wider appreciation and understanding of geoheritage sites and geosciences.

Over the last 10–15 years, various creative approaches have been developed to enhance geoscience communication, including information panels (Bruno and Wallace, 2019; Pasquaré Mariotto and Venturini, 2017), 3D models/virtual site visits (Dolphin et al., 2019; Hoblea et al., 2014), animations (Lansigu et al., 2014), poetry (Illingworth, 2023), mobile applications (Cayla, 2014), soundscapes (Connor and Maculuve, 2025) and dance (Matias et al., 2020). Regardless of the approach, many studies share the common theme of storytelling (Migoń and Pijet-Migoń, 2017; Van Loon et al., 2020; Matias et al., 2020; Illingworth, 2023; Stewart and Nield, 2013), praising its utility in delivering information in ways that connects with the audience cognitively and emotionally (Dahlstrom and Scheufele, 2018).

In analysing these initiatives, only a few studies have conducted quantitative or mixed-method assessments of the impacts of interventions. One of the few examples is Mani et al. (2016), whose study found a 12 %–17 % improvement in participants' knowledge on volcanic hazard after playing a video game on the topic. Such impact assessments demonstrate the potential of geoscience communication, but these studies usually evaluate short-term effects, from data collected immediately after interventions. Given the goal of geoscience communication to increase public discussion and sustained attention to geosciences (Illingworth et al., 2018), longer-term assessments are necessary to provide a more comprehensive understanding of the communication outcomes.

Videos have been used in scientific communication studies across various fields, including ecology (Ruzi et al., 2021), climate science (Ettinger et al., 2021), and medical sciences (Dudley et al., 2023). Videos offer several advantages as a mode of communication. Delivering information in person or through videos and other visual media outperformed written text in 15 studies (Dudley et al., 2023). Videos facilitate enhanced emotional engagement of the audience (Yadav et al., 2011) and easier comprehension through the combination of narration and images (Moreno and Mayer, 1999).

Recognising the importance of videos in science communication, studies have provided guidelines for developing effective videos. Desired features included visual attractiveness, brevity, engaging content, and consideration of the audience needs and expectations (León and Bourk, 2018). The optimal length of educational videos has been identified as around 6–9 min (Guo et al., 2014; Kaim et al., 2020; Manasrah et al., 2021). Guo et al. (2014), found that when videos exceeded 9 min, viewers are only focused on the content for less than 50 % of the total duration.

In experimental settings, videos are commonly used to assess changes in opinions or attitudes after viewing (Kaim et al., 2020; Dunn et al., 2020), as well as to compare the effectiveness of different science communication strategies. For example, Ettinger et al. (2021) used short videos to compare differences in viewers' reactions to videos with a pessimistic or an optimistic outlook on the trajectory of climate change. In geoscience communication, however, the utility of videos has seldom been studied quantitatively. One notable exception is Zawacki et al. (2022), who analysed the relative success of 48 geoscientific Tik Tok videos. This study found that while the most viewed videos were short clips related to news events; it was the longer videos, with fewer views, that garnered the most user engagement through comments and “likes”. Given the demonstrated efficacy of videos in communicating science and technology, there are further opportunities to test their applicability for geoscientific topics, particularly in areas of the world where geoscience communication is in its infancy.

As a preparatory phase to determine the most appropriate content of the videos, focus group meetings were conducted in November 2022 at Buhais Geology Park (UAE), and in December 2022 at the Oman National Museum. Participants are described in Table 2. The authors presented the study aim, design, target audience, and focus to the participants, followed by descriptions of the relationship between archaeology, culture, and palaeoenvironment.

Through deductive and inductive content analysis (Bengtsson, 2016) via NVivo, five key themes were identified as below:

egin the video with visually engaging footage and familiar sites before introducing technical Quaternary elements;

Based on the insights above, two videos were developed by the authors in collaboration with partners. The videos were designed to be of similar length, with the primary distinction being an emphasis on the geocultural context in one (referred to as the geocultural or treatment video), and the exclusion of the geocultural context in favour of a more detailed description of the Quaternary science in the other (Quaternary science or control video). Other elements were designed to be as similar as possible. The videos incorporated footage filmed by the authors, as well as videos and photographs provided by collaborators, along with Creative Commons licensed visuals and music. Each video featured voice narration in English, complemented by dialogues with experts conducted in Arabic, with corresponding translation provided as subtitles in Arabic or in English. The videos were created specifically for this research project and are not intended to function as comprehensive museum or geopark interpretation tools. Prior to distribution, both videos were piloted and reviewed by focus group participants to ensure scientific accuracy and appropriate representation of cultural and social content.

The videos are available online (Control Video: https://ora.ox.ac.uk/objects/uuid:e14384cb-68fa-45f1-914d-3204e05bb3ea, last access: 20 June 2026, Treatment Video: https://ora.ox.ac.uk/objects/uuid:ee8c3f2a-1888-4967-a104-b5799ba68d8d, last access: 20 June 2026), and their structure, duration, and differences are summarised in Table 3.

The total duration of the videos falls at the upper end of the 6–9 min duration range recommended in the literature for optimal audience engagement, at 9 min 5 s (545 s) for the geocultural video, and 8 min 56 s (536 s) for the Quaternary science video. The Quaternary science video incorporated additional scientific detail. The total duration of geocultural content was 140 s (approximately 26 %) in the geocultural video.

In these videos, three case study sites were selected to introduce four key Quaternary geoheritage sites in the region. These sites were selected for primarily for their scientific importance, but also for the representation of the variety of Quaternary sites in southeast Arabia. Table 4 summarizes the key features of the selected sites.

In phase 3, an online experiment was conducted using the videos produced in the second phase. Participants were recruited based on the inclusion criterion of being residents of the UAE or Oman. Recruitment efforts included calls at universities, as well as emails and posts through nature societies, museums, and social media platforms. The experiment consisted of three waves of surveys, including a questionnaire administered before, immediately after, and three months after watching the video. Upon completion of the pre-video questionnaire, participants were randomly divided into two groups: Group A (the treatment group), who watched the geocultural video, and Group B (the control group), who watched the Quaternary science video.

The study received 160 valid responses in Wave 1 (pre-video), 104 in Wave 2 (post-video), and 82 in Wave 3 (3 months after). Participant characteristics are summarised in Table 5. Although nationals of UAE and Oman were overrepresented relative to national demographics, comparative analyses revealed no statistically significant differences between national and non-national participants. While this study originally aimed to target the general public, the final sample was skewed towards a young, highly educated, and environmentally engaged demographic. This profile aligns with audiences commonly identified as potential geotourism participants in previous studies (Dowling and Allan, 2018; Kim et al., 2008; Štrba, 2019).

Drop-out rates were within or better than the expected range of 30 %–40 % for longitudinal studies (Grønmo, 2019), with a particularly low drop-out rate of 21 % between Waves 2 and 3. Given the total population of southeast Arabia at 14.1 million, according to Yamane's formula for sample size (Yamane, 1967), 104 participants who completed until Wave 2 represent the population of this region at a margin of error at ±9.8 %, while 82 participants who completed all three waves have a ±10.9 % margin. While smaller than the sample required for a ±5 % margin, this size was considered adequate for exploratory analysis, especially given the strong caveat that the study focused on a specific population of university students and adults with pro-nature views.

The majority of participants were university students (n=78), including 47 studying geosciences and 31 studying other disciplines. Of the 23 employed participants, only one worked in geology-related field. Participants who are geoscience students or professionals were classified as domain specialists (n=48) and the rest were categorised as non-specialists (n=56). Before watching the video, specialists were significantly more likely to report awareness of geoheritage sites (t test p<0.001) or Quaternary geoheritage sites (t test p=0.051 in χ2 test and p value = 0.035 in Fisher's exact test). Many specialists knew some geoheritage sites (95 %), but only half of them (50 %) could recognise or identify Quaternary geoheritage sites.

Participants of this study, regardless of their demographics, expressed strong pro-conservation attitudes prior to viewing the videos, with high rating of the importance of protecting natural heritage (mean = 9.2 out of 10, SD = 1.4). Support was similarly strong for geoheritage sites (mean = 9.0, SD = 1.7) and Quaternary geoheritage sites (mean = 8.6, SD = 2.0), although increasing specificity was associated with greater variability in responses.

All parts of the experiment were conducted using online questionnaires, with options to answer in Arabic or English. The questionnaire was first prepared in English and then translated into Arabic by one of the authors (Husam Al Rawahi), a native Arabic speaker and a geoscientist. Differences between the two language versions were reviewed and considered negligible in analysing the answers.

The questionnaires assessed participants' knowledge and interest in geoheritage sites, Quaternary geoheritage sites, archaeological sites, and the relationship between Quaternary geoheritage sites and archaeological sites. The Quaternary geoheritage section included two sets of knowledge test questions: one in which participants were asked to select photos of Quaternary geoheritage sites from a set of photos (see Fig. 3), and another in which they were asked to choose correct statements on relationships between Quaternary geology and archaeology observable in southeast Arabia (see Supplement 2 for full question wording). Both knowledge test questions were presented as multiple-choice questions with multiple correct and incorrect answers, to which participants were asked to select all correct answers. The score for these questions was calculated as the number of correct answers selected, minus the number of incorrect answers selected. After the Wave 2 questionnaire, participants received a list of additional resources where they could learn more about geoheritage sites and sites that were featured in the video. In Wave 3, additional questions were included to assess participants' impressions of the video and its impact.

Each questionnaire was designed to take between 5–10 min to complete. Most questions utilised a five-point Likert-like scale, in which higher values reflect higher levels of agreement, interest, or perceived importance, with exceptions including Yes/No questions, text-entry questions (mostly in Wave 3 regarding impressions and impact of videos), multiple-answer questions (including knowledge test questions), and questions on the importance of protecting certain types of sites, which were presented as a 10-point slider-scale. Text-entry questions were included to assess participants' recollection of Quaternary geoheritage sites.

The data collected from the questionnaires underwent quantitative and qualitative analysis, based on the nature of the questions. Although some discussion continues on the treatment of aggregated Likert-type data as an ordinal scale rather than discrete categories (Bishop and Herron, 2015), this study employed parametric methods to analyse this data in line with other studies such as De Winter and Dodou (2010).

A qualitative assessment was conducted on text entry question responses involving deductive and inductive content analysis using NVivo to categorise themes related to the research questions and emerging themes from the responses. The predetermined themes used for the deductive content analysis included those related to the key points mentioned in the focus groups, and options indicated in the questionnaire as multiple-choice questions.

Statistical analyses followed methods commonly used in Randomly Controlled Trial (RCT) experiments, where both treatment and control groups are present. Since both groups watched a video on geoheritage which would impact the outcome of the post-treatment (Wave 2) and follow-up (Wave 3) surveys, the study opted to use difference-in-differences (DiD) analysis with fixed effects for the treatment group and the survey wave. DiD analysis is a specific type of fixed effect analysis that is commonly used to evaluate treatment effects in natural experiments (see Abadie, 2005 for a detailed discussion). In this study, DiD enables an analysis of changes in participants' responses over time in the treatment group vis à vis the changes in the control group, thereby isolating the effect of including geocultural information in the video.

For Waves 2 and 3, all responses were initially compared to responses in Wave 1 using a t-test to account for statistically significant differences. Subsequently, DiD was conducted to assess the effects of the treatment. DiD analysis was conducted at two levels. First, it was conducted for all 82 participants who completed Wave 3. Then the participants were subdivided into specialist and non-specialist respondents to examine whether the treatment effect varied based on the respondent's level of prior subject knowledge.

When applied to samples of the size used in this study, estimates of treatment effects are accompanied by substantial statistical uncertainty. With 82 complete cases, the study is underpowered to reliably detect small effects, particularly under inference procedures that account for within-respondent dependence over time. We therefore emphasize effect-size estimates and uncertainty intervals by reporting 95 % confidence intervals (CIs) for our DiD analyses. Confidence intervals allow interpretation of the range of plausible directions and magnitudes for the true effect sizes, which is more informative for these data than significance-level reporting alone. For example, in cases where confidence intervals are narrow and centred near zero, we can rule out the presence of large treatment effects even when results are not statistically significant in the conventional sense. Across outcomes, many estimates remain imprecise, and we interpret findings in light of this uncertainty as reflected in the reported confidence intervals.

Survey responses indicated positive and sustained impressions of both videos. Approximately half the participants in each group reported watching the video more than twice. Self-reported memory of content was higher in the treatment group (25 %) than in the control group (16 %).

Most participants (80 %) considered the video length appropriate, although 17 % felt it was slightly too long. This perception was more common in the control group (24 %) than in the treatment group (13 %). Regarding impact, 91 % of respondents reported increased interest in Quaternary geoheritage, with a slightly higher proportion in the treatment group, though differences were not statistically significant.

Of the 82 Wave 3 respondents, 46 provided open-ended text feedback (20 from treatment group and 26 from control group). These comments frequently cited positive aspects including learning outcomes (13), clarity of presentation (8), and beautiful scenery (8). Participants described the videos as accessible, clear, and engaging, and several highlighted the dialogue with local geologists as particularly effective. In contrast, only a small number of comments explicitly referenced Quaternary geoheritage sites (6), as opposed to general geoheritage sites (19). Suggestions for improvement were provided by 20 respondents, most commonly recommending a shorter duration, predominantly from control group participants (7 out of 8 instances). Conversely, several treatment group participants suggested including additional content to further expand their knowledge. Other recommendations focused on points to improve the production quality, such as including short animations (2), improving the narration (2), and improving the overall quality (3) as well as the delivery of the content, such as reducing complexity (3), and making the content more interesting (5).

Statistical analysis of the questionnaire answers focused on the questions pertinent to Quaternary geoheritage sites and the geocultural context. The results of the seven target questions are shown in Fig. 4 and Table 6 and analysed further in the following sections. The detailed statistical figures for the DiD analysis can be found in Supplement 3.

Interest in Quaternary geoheritage increased significantly immediately after viewing (t test p=0.0011), but this uplift was not evident at the three-month follow-up when comparing pre-video and follow-up distributions (p=0.55). Perceived knowledge, by contrast, increased strongly in the post-treatment survey (p<0.001) and remained significantly higher than baseline at follow-up (p<0.001), despite some degree of regression towards the mean.

Comparative analysis between the control and treatment group shows that for interest, DiD coefficients were small with confidence intervals around zero, effectively ruling out large effects. The only exception was a possible small positive effect among non-specialists at the follow-up survey, albeit with substantial uncertainty. For perceived knowledge, however, DiD analysis suggested a negative effect associated with assignment to the treatment group after Wave 2 for both specialists and non-specialists, although this effect dissipates by Wave 3. When comparing specialists and non-specialists in terms of perceived knowledge in Wave 2, specialists reported a greater increase than non-specialists.

The analysis of the knowledge test question on Quaternary geoheritage identification revealed a highly significant improvement in the scores immediately after video viewing (p<0.001). While a substantial degree of mean reversion was observed by the time of the follow-up survey, scores remained significantly higher than the pre-video scores (p=0.013). The DiD analysis provided a clearer picture. Immediately after viewing the videos, the treatment group showed smaller improvements than the control group. This initial better performance by the control group was driven entirely by specialist respondents. However, by the follow-up survey, the two groups performed almost the same, with the treatment group scoring slightly higher, despite the lower score in Wave 1.

Interest in the relationship between Quaternary geology and archaeology showed a small but statistically significant increase immediately after viewing (p=0.028) at the level of all participants, but was statistically indistinguishable from baseline three months after watching the video. On the other hand, self-reported knowledge increased significantly immediately after viewing (p<0.001) and remained significantly higher than baseline at follow-up (p<0.001), indicating that participants' perceived understanding of the geocultural relationship continued to develop over time.

DiD analysis suggests that the treatment–control contrast was clearer for interest than for perceived knowledge. For interest, the treatment group, especially the non-specialists, showed a notably greater increase than the control group both immediately after viewing the video and three months after, with the follow-up effect particularly strong (CI lower-end value above 0), indicating a conventionally statistically significant finding, despite the small sample size.

Overall scores on the knowledge test for the relationship between Quaternary geology and archaeology question were extremely low across all survey waves, with mean values hovering around zero. Nevertheless, a statistically significant improvement was observed immediately after viewing the videos (p=0.0014). Scores in the follow-up survey were higher than baseline but not significantly so at the 95 % confidence level (p=0.109). The DiD analysis suggests that, overall, there was little to no difference between the treatment and control group.

A statistically significant increase in the view towards the protection of Quaternary sites was observed immediately after viewing the videos (p<0.001). Although scores exhibited a high degree of regression towards the mean by the follow-up survey, they remained slightly higher than pre-video levels. However, this difference was not statistically significant (p=0.142). It is notable that baseline scores were already high (mean ∼8.5, with the highest possible rating being 10), and very few respondents rated these sites as a low priority for protection.

DiD analysis suggests that the positive gains in this question were consistently stronger for the treatment group, with greater increases in protection priority than those in the control group overall. This effect was evident both immediately after and three-months after viewing the video. The positive effect was particularly strong for non-specialist respondents with the treatment group in particular. Specialists entered the study with higher baseline scores (a mean value of 9.13 versus 8.23 for non-specialists), which could have constrained the magnitude of observable change due to ceiling effects.