The questionnaire (see the Supplement) contained seven questions, with each question designed to evaluate students' learning of a topic that was explored in selected videos. To elicit a wide range of responses from the students, we used open-ended questions (e.g. “what are the causes of earthquakes?”), drawing strategies (e.g. “draw the Earth's interior”) and analysis of photographs (e.g. “can you identify non-structural hazards in each photo?”). We also included questions requiring “yes” or “no” answers (e.g. “can earthquakes be predicted?”). For these questions, we included “I don't know” and “other (please specify)” in the answers from which to select. The only demographic information collected was student gender. To anonymously link the pre- with post-assessment data collected from each individual student, we asked students to create a confidential identifier unique to themselves and write it on both pre- and post-assessment questionnaires. Using this method, we were able to collect data from a total of 77 students from Tajikistan and the UK.

Students' written responses to survey questionnaires were first categorized into appropriate groups, based on the individual response to each question. We then assigned a score value to each response group, indicating the level of understanding associated with the response. For example, students who mentioned volcanoes and mountains in their answers as the primary cause of earthquakes received a score of 1; those mentioning plate tectonics received a score of 2; those with no or irrelevant answers received a score of 0. Similarly, students' graphical responses (question 6 in the survey) were analysed using the evaluation rubric of Steer et al. (2005). For example, students' drawings of the Earth's interior were scored 0 (no conceptual framework) to 5 (advanced understanding). Students' scored responses to pre- and post-assessment questions were then analysed for comparison using a two-sample t test when data followed the normal probability distribution and the non-parametric Kolmogorov–Smirnov (KS) test when the probability distributions were non-normal. For survey questions that required binary answers (“yes” or “no” responses, e.g. in questions 3–4 in the survey), we used the McNemar test to compare results. The significance level (alpha value) was set to 0.05, and results were considered statistically significant if p<0.05.

We selected three videos for classroom testing in the UK and Tajikistan. These videos included the first two earthquake science video lessons in the series (i.e. Earth's interior and plate boundaries) and the last earthquake hazard and safety video (i.e. non-structural hazards; see Fig. 3). These videos were chosen since they required no previous knowledge of earthquakes, and they cover the fundamental concepts related to earthquakes (i.e. Earth's interior and plate tectonics) and the most common cause of earthquake-related injuries (i.e. non-structural hazards). Furthermore, these videos use a wide range of pedagogical approaches to teaching and learning. For example, the Earth's interior video follows a model-based, conceptual change approach to teaching to improve students' understanding of the Earth's structure, while the plate boundaries video is data driven and follows the jigsaw method of cooperative learning (i.e. students depend on each other to succeed), as shown in Sawyer et al. (2005). The non-structural hazards video uses a place-based approach, promoting learning that is rooted in the students' own place (i.e. their classroom) to raise awareness about non-structural hazards and mitigation measures in school classrooms where the videos were tested. This range of teaching methods tested allows us to gain the most information about students' diverse and complex individual learning needs, in response to paired teaching, from the information collected in the questionnaires.

The three aforementioned videos were tested with 38 sixth-grade students (12 years of age; 50 % female) and 39 ninth-grade students (12–14 years of age; 42 % female) from two school classes in Dushanbe (Tajikistan) and London (United Kingdom), respectively. The school in Dushanbe (the capital city of Tajikistan) is a public school located in the city centre and was selected for this study by the Institute of Geology, Earthquake Engineering, and Seismology of Tajikistan because of its previous collaboration with Mohadjer et al. (2010). The school was recently built in compliance with the existing seismic building codes of Tajikistan. Interactive whiteboards are present in many of its classrooms, and there is a new computer lab with individual workstations. The video testing was conducted over 5 d during school hours (i.e. 08:00–12:00 local time – LT) by the lead author, with assistance from local teachers. The local language (Tajik) was used for teaching and in all written and multimedia materials, including the videos which were dubbed into Tajik. Resources needed for testing (e.g. maps, Slinky toys, and play dough) were provided by the lead author. The school in London (United Kingdom) is a secondary (11–18 years of age) academy located in the heart of the city. The London school was selected through our existing teachers' network in the UK. The videos were tested by two geography teachers with 1–3 years of teaching experience over a testing period of approximately 50 d. The difference in the testing periods between the two schools was to accommodate different schedules teachers had to follow. While the UK teachers could spread out the video lessons across a 50 d period, the lead author of this paper had to follow a restricted schedule of 5 d for video testing in Tajikistan.

Students' understanding of the Earth's interior before and after the classroom testing of the Earth's Interior and Plate Tectonics video is shown in Fig. 4. While the majority of the Tajik students' responses, both before and after watching the video, show no to little understanding of the Earth's interior (scores of 0–2), responses given by the UK students are more evenly distributed between no or little understanding and higher levels of understanding (scores of >2). Both groups, however, show an increase in their understanding of the Earth's interior after watching the video. These observations are supported by statistical analysis of the results. More specifically, 74 % of students (28 students) from Tajikistan and 48 % from the UK (19 students) demonstrated having a no or a basic conceptual framework about the Earth's interior (scored 0–1) prior to video testing. After video testing, a large percentage of Tajik and UK students (58 % and 52 %, respectively) demonstrated an increased level of understanding of the Earth's interior (scored 3 or higher). The difference between Tajik students' responses before and after video testing was statistically significant (above 95 % level), using the Kolmogorov–Smirnov (KS) test (D stat – 0.31; D crit – 0.30). Similarly, the difference between UK and Tajik students' responses before and after video testing was significant (above 95 %; D stat – 0.33; D crit – 0.30).

We grouped students' responses into six categories (Fig. 5). In their responses to “What are the causes of earthquakes?”, 90 % (35 students) of the UK students mentioned plate tectonics, while 46 % (17 students) of the Tajik students made references to mountains and volcanoes (with only 1 out of 38 students mentioning plate tectonics) before video testing. After video testing, the Tajik students showed little improvement in their understanding of the causes of earthquakes. The difference between UK and Tajik students' responses prior to video testing and their responses afterwards was significant, above the 95 % level, using the KS test (D stat – 0.84 and 0.79; D crit – 0.30 and 0.30).

Figure 6 shows the students' ability to identify non-structural hazards in example photographs. Non-structural earthquake hazards are caused by the furnishings and non-structural elements of a building (e.g. suspended ceilings and windows). In general, students from both groups identified non-structural hazards that are located above the ground (e.g. hanging television set and plant pots stored above cabinets) and missed those near the floor (e.g. desks and chairs). Both groups demonstrated some knowledge of non-structural hazards found in typical school classrooms prior to video testing and showed some improvement after video testing. However, only the difference between pre- and post-assessment responses by the UK students was statistically significant above the 95 % level, as indicated by the KS test (D stat – 0.43; D crit – 0.30).

Figure 7a shows students' responses to “Is it possible to know the exact timing of earthquakes before they occur?” While the majority of Tajik students indicate “no” in their responses before and after video watching (55 % and 68 %, respectively), UK students' responses are more evenly distributed between three categories of “no”, “sometimes”, and “I don't know” prior to video watching (36 %, 33 %, and 28 %, respectively) as well as after video testing (46 % indicate “no”; 38 % indicate “sometimes”). After video watching, a notable percentage of Tajik students (21 %; 8 students) believes that it is possible to know the exact timing of an earthquake.

In their pre-assessment responses to whether it is possible to know where an earthquake can occur (Fig. 7b), a significant percentage of both groups (47 % – Tajik; 41 % – UK) believe that is sometimes possible. After video watching, Tajik students' responses are divided almost evenly between the four different categories as follows: 26 % (10 students) indicate that it is possible to know where earthquakes can occur (“yes” response), while 18 % (7 students) believe otherwise (“no” response); 21 % (8 students) indicate it is sometimes possible, and 26 % (10 students) give irrelevant answers. In contrast, the majority of the UK students' post-assessment answers are “yes” to earthquake location prediction (62 %; 24 students), with “sometimes” indicated in 28 % (11 students) of responses. The differences between UK pre- and post-assessment responses is statistically significant above the 95 % level, as indicated by the McNemar test (chi stat – 6.66; chi crit – 3.84).

Additionally, we asked students the following open-ended question: where in the world do earthquakes occur most often? Based on their answers, we created five response categories (Fig. 7c). A large percentage of students' responses from both groups include country names before and after watching the videos. UK students also included plate boundaries in their responses both before and after watching videos (21 % and 40 %, respectively). In contrast, Tajik students made references to mountains or volcanoes (39 % and 44 %, respectively). These observations match students' responses to the causes of earthquakes (Fig. 5). The differences between UK and Tajik post-assessment responses is statistically significant above the 95 % level, as indicated by the KS test (D stat – 0.38; D crit – 0.30).

While a notable percentage of Tajik students' responses (50 %, 19 students) indicate an increase of at least one score point in their understanding of the Earth's interior, only four (out of 38) students demonstrated an advanced understanding of the Earth's interior structure where scale is important (score 4–5). This is despite the fact that students were shown a diagram of a cross section of the Earth (with all layers drawn to scale and labelled) by the video teacher and participated in a classroom activity during which they compared the interior structure of a hard-boiled egg with that of the Earth. These concepts, however, were unfamiliar to Tajik students prior to video testing and were not repeatedly reinforced during and after video implementation. It is important to note that more than 50 % of Tajik students did not label the Earth's layers in both pre- and post-assessment surveys since this was not specified by the question. However, the post-assessment data show a 24 % increase in the number of students who drew the interior of the Earth as concentric circles. Therefore, it is possible that students' understanding of the Earth's interior after video watching is underestimated. For the UK group, 46 % (18 out of 39 students) showed no or a basic conceptual framework of the Earth's interior after video testing, with only 31 % (12 students) showing some improvement. We, therefore, conclude that diagrams and simple analogies can bring some improvement to learning these concepts, but to make significant conceptual advances, concepts must be revisited and reinforced repeatedly.

The Earth's interior and plate boundary videos did not increase students' understanding of earthquakes and associated processes. After video testing, only one Tajik student (out of 38) listed plate tectonics as the main cause of earthquakes. The relationship between plate tectonics and earthquakes was briefly mentioned by the video teacher in the Earth's interior video. However, during the testing of the plate boundary video, students used different data sets (including an earthquake map) to observe data behaviour near or at plate boundaries. It is possible that the students' incomplete understanding of the Earth's interior structure (as described above) hindered their learning process associated with earthquakes, as shown by Barrow and Haskins (1996). Another hindering factor could be the lack of previous exposure and concept reinforcement during or after video testing. Relating earthquakes to volcanoes and mountains is part of students' pre-existing knowledge, as demonstrated in 39 % to 50 % of students' pre-responses to question 1 (i.e. “what are the causes of earthquakes?”) and question 4 (i.e. “where in the world do earthquakes occur most often?”), respectively. Students' pre-existing knowledge of why and where earthquakes occur was revisited and reinforced in the plate boundary video and the accompanying classroom activities, where students explored relations between distribution of volcanoes, topography, and earthquakes. Relating these processes to plate boundaries, however, was a new concept that was not reinforced. Unlike previous work by Mohadjer et al. (2010), we documented no mention of myths, legends, or religious explanations in Tajik students' responses to causes of earthquakes. This could be due to (1) different data collection methods (anonymous survey questionnaires in this study vs. individual or group interviews in Mohadjer et al., 2010) and (2) differences in students' ages or grade level (the 2010 participants were 2–3 years older than the 2018 participants), and/or (3) this may be a reflection of changing earthquake perceptions over the last decade. In contrast, nearly all UK students (35 students) connect earthquake occurrence with plate tectonics prior to video watching. Because of their previous knowledge, it is difficult to assess the effectiveness of the plate boundary curricular activities conducted with this group.

The two closed questions in the survey that assessed the students' views of earthquake prediction (questions 2 and 3) in terms of earthquake location and timing were not directly addressed by our curricular material. However, these questions were selected to (1) assess the students' current perception of and views on earthquake prediction and forecasting in general and (2) to find out whether learning about the Earth's interior and earthquake-related processes alone can change the students' views of earthquake prediction and forecasting. This information is important, since earthquake prediction or forecast, or lack thereof, may influence preparedness attitudes and behaviours in some communities. Prior to video watching, approximately 60 % (23 Tajik students) said “no” to earthquake prediction (in terms of date and time) and “yes” or “sometimes” to earthquake forecasting (in terms of location). In contrast, these values for the UK group are lower (36 %; 14 students) and higher (77 %; 30 students), respectively. The only significant difference (95 % level) between the students' pre- and post-answers was observed for the UK group for earthquake location forecasting. This may indicate that our curricular material (particularly the plate boundary classroom activity), which emphasizes that earthquake locations are not random, was effective in changing students' understanding of earthquake location forecasting. A notable portion of Tajik students (26 %; 10 students), however, misinterpreted the survey question, as is evident in their irrelevant responses to earthquake location forecasting (e.g. listing streets and schools as locations of earthquakes).

For both student groups, our non-structural hazard video increased the students' ability to identify non-structural hazards in example classroom photos. However, the increase in hazard identification is significant (95 % level) only for the UK group. This could be due to lack of exposure to earthquake shaking during which non-structural elements of a building can pose hazards to the building's occupants. Having experienced earthquakes and being familiar with some hazards, Tajik students' give similar responses before and after video watching.

Both teachers found the curricular material to be presented clearly, to be appropriately geared to the level of their students, and to be valuable in helping students understand and learn the lesson content. However, there were differences in how they rated the materials in terms of relevance. In general, classroom activities carried out during video breaks were described as relevant by both teachers. However, both teachers found activities related to the cost-benefit analysis of non-structural hazards in video 3 to be irrelevant with low levels of student engagement when compared with other assignments in their class. One teacher explained, “[students] did not find the tasks that useful/applicable to life as we are not in an earthquake-prone area.” Similarly, another teacher said, “[…] compared to the way we normally teach tectonic hazards, this (1 h 40 min) felt like a lot of curriculum time on a relatively narrow aspect of the topic.” There were some differences between teachers' opinions and experiences with video testing. For example, while one teacher found the egg analogy to be a “very relevant” activity and described students' level of engagement with this activity as being “a great deal more” than with other classroom assignments, the other teacher recommended omitting this activity, saying “[…] I think they could have completed this task more effectively without being given and cutting up the egg (which took up time and [was] a distraction).” Teachers played the entire videos in the classroom, with the exception of the plate boundary video for which one teacher played only two out of five video segments and skipped the related classroom activities. The same teacher described students' level of engagement to be below average for the two activities that were carried out.

Video testing in Tajikistan had to be conducted by the lead author (a geosciences researcher and teacher) due to the local teachers' lack of interest and/or confidence in having the required teaching skills and knowledge. This was despite making teaching materials accessible and available in the local language in advance of video testing and holding an informal meeting with local teachers during which they were introduced to paired teaching and lesson content. To further support and encourage the local teachers to participate in the video testing, a trained teaching assistant familiar with the paired-teaching videos was made available. The school principal was supportive of teachers' participation but did not require it. Due to limited time and resources, we had to carry out the video testing with little contribution from local teachers. The teacher feedback discussed here, therefore, reflects our experience with video testing.

With the exception of the plate boundary video, the level of content covered in both the Earth's interior and non-structural hazards videos were appropriate for Tajik students. During the plate boundary video testing, most students were not only challenged by the lesson content (e.g. understanding maps, data relationships, and classification) but also struggled with understanding and following the pedagogical approach (e.g. collaborative learning) required for classroom activities. This lowered the class pace, and therefore, the video was played partially, with some classroom activities omitted. Students' level of engagement was the lowest for the plate boundary video and the highest for the non-structural hazard video, which focused on a topic most student could relate to (identifying hazards in their own classroom). The level of student excitement with lessons, however, was the highest when watching the Earth's interior video, which was taught by a UK-based video teacher. The excitement was less for the subsequent videos, which were taught by the lead author who had to act as both the video teacher and the in-class teacher.

The video testing was affected by both the classroom testing period and the choice of space and place for curriculum implementation. For example, the Earth's interior video was tested during normal school hours but outside the classroom environment in Tajikistan. For this lesson, the stage in the school theatre was selected by school authorities because the equipment required (e.g. projector, screen, and sound system) was only available for use there. For the remaining two videos, the teaching space was changed to a regular classroom and a computer lab. While the latter restricted the students' movements and group formation for activities, the former provided a familiar and flexible space where the students and the in-class teacher could easily rearrange chairs and tables according to their needs. In addition, it is possible that the differences in the testing periods between the UK and Tajikistan groups (50 d vs. 5 d, respectively) influenced the study results. For example, due to time constraints, some in-class activities had to be shortened or skipped in Tajik classrooms. That, combined with technological issues, posed additional challenges to video testing by shortening video testing period. Teacher feedback highlights some important differences and similarities during classroom video implementation in two geographically and culturally different parts of the world. As the above observations show, the students' experience with earthquakes and related hazards influences their level of engagement with lesson content. While technological shortcomings and ineffective classroom space and management significantly challenged video testing in Tajikistan, these factors were non-existent in the UK school. The poor classroom management in the Tajik school was exacerbated by the lack of the local teachers' involvement, which led them to bring in a new teacher (lead author) who was unfamiliar with the specifics of the classroom culture. Despite these differences, both the UK and Tajik teachers acted flexibly with video testing by modifying the video lessons according to the needs of their students and their local environment (e.g. skipping video segments irrelevant to students' lives or shifting to printed lesson plans and other resources when technology failed). To maximize the impact of our paired-teaching approach to earthquake education, we suggest exercising flexibility when using our videos and contextualizing video content and learning activities to increase their relevance. To encourage and assist teachers with lesson preparation, we plan to improve our teacher segment by creating a guide for each video, in a printable format, with descriptions of activities covered in each segment, as requested by UK teachers.

An effective DRR-related, school-based education is one that is contextualized to meet local needs and carried out in the cultural context that surrounds the implementation of the curriculum. Students' and teacher's needs and goals, local constraints, and schools' pedagogical values are some of the factors that shape a classroom culture. While some topics, such as the Earth's interior and plate tectonics are often covered to varying degrees in most schools around the world, knowing how to identify and fix non-structural hazards related to earthquakes might not be as relevant to schools located in non-seismically active regions. Similarly, content taught using pedagogical approaches that are unfamiliar to some teachers and students can hinder effective learning. For example, most of the learning activities in this study were based on cooperative learning (e.g. jigsaw concept) and involved group discussions, where everyone's input was encouraged. Because they were unfamiliar with this approach, these activities appeared to be unstructured to most Tajik students, leaving some students being uncomfortable with sharing their opinions.

The low level of engagement by local teachers in Tajikistan, with respect to serving as in-class teachers in the paired-teaching approach, may be due to their unfamiliarity and discomfort with collaborative learning methods and the use of video technology. Since the paired-teaching video lessons were designed to be a complete resource (i.e. containing video segments, teacher's guides, downloadable handouts, and lists of other resources relevant to the topic), no teacher training was provided for using these videos. However, teachers were encouraged to view the videos and familiarize themselves with the content before using them in their classrooms. This study, however, reveals that these videos may not be seen as a complete resource by some teachers. While the UK teachers tested the videos with minimal input from video creators, teachers in Tajikistan asked to observe classroom testing of the videos. This request was made despite the fact that the teachers were offered training in delivering the videos and/or the option to co-teach the video lessons with experienced instructors. Similar to teachers in Tajikistan, teachers in China, Japan, and Malaysia, where rote learning dominates classroom culture, experienced difficulties with paired teaching (Larson and Murray, 2017). Therefore, the textbook-based classroom culture may partly explain why Tajik teachers did not want to actively engage in video testing. In addition, teachers' low level of technology acceptance and readiness for teaching and learning has been shown to hinder their engagement with technology-based pedagogical approaches (Shukor et al., 2018). Our study, therefore, shows that the paired-teaching pedagogy is not a one-size-fits-all teaching approach and depends on the classroom culture and teacher's comfort operating within it. Thus, when developing curricular material, both teachers' and students' involvement are key to ensure an appropriate selection of content and pedagogical approaches. This can be achieved through informal classroom observation and discussions of goals and pedagogical expectations with classroom teachers and students, as well as providing ongoing, high-quality pedagogical training that supports teachers in adopting a more student-centred and collaborative teaching style for their classrooms.

The importance of public engagement activities is increasingly recognized by scientists, funding institutions and policymakers (NSF, 2015; Rauws, 2015; European Union, 2002). However, many of those who practice science education and outreach do not always evaluate their work rigorously, and even fewer publish and share their results in peer-reviewed journals. By testing our videos in school classrooms located in different countries (UK and Tajikistan), we were able to assess the effectiveness of our educational materials and identify potential factors that influence learning. However, the assessment of our evaluation strategy revealed several issues. While using questionnaires with students may be a time and/or cost-efficient method of collecting information anonymously, if not designed and explained carefully, the students may interpret the questions differently. This was the case with some students in Tajikistan, where they perceived the questionnaire as an exam and strived for correct answers as opposed to freely sharing their existing knowledge. Despite gaining new knowledge, some Tajik students gave identical responses to pre- and post-assessment questions in order to be consistent as opposed to being correct. We, therefore, recommend using questionnaires as one of several methods for collecting and evaluating data. Strategies such as conducting face-to-face interviews with students or arranging for and recording group interviews can minimize misunderstanding and provide important insights into student–student and student–teacher interactions that enhance or hinder learning. Thus, an effective evaluation strategy should consider students' familiarity with data collection methods to ensure that the students understand what they are being asked to do and why.