Citizen science and the United Nations Sendai Framework

The United Nations (UN) Sendai Framework for Disaster Risk Reduction 2015–2030 (SFDRR) was signed on 18 March 2015, at the UN World Conference on Disaster Reduction, by 187 member states. Its general aim is to prevent and reduce disaster risks by promoting good governance in disaster risk reduction strategies, specifically in developing regions (Kelman and Glantz 2015). To achieve this goal, the SFDRR defines priorities for action (article 20) including (1) “understanding disaster risk” and (2) “strengthening disaster risk governance to manage disaster risk.” To help achieve these objectives, the SFDRR also defines thirteen guiding principles (article 19). Two of them explicitly focus on citizen engagement, by highlighting that: (i) “disaster risk reduction requires an all-of-society engagement and partnership” and (ii) “disaster risk reduction requires a multi-hazard approach and inclusive risk-informed decision-making.

The coupling of priorities (1) and (2) and the guiding principles (i) and (ii) points to citizen participation in scientific research and expertise as a tool to study and collectively manage disaster risks. For this reason, the role of citizen science for disaster risk reduction has been discussed at length in the literature since the adoption of the SFDRR (Paul et al. 2018). As shown in different case studies, citizen science could be used within all five steps defined by Moe and Pathranarakul (2006) for disaster risk management: prediction, warning, emergency relief, rehabilitation, and reconstruction (e.g., Marchezini et al. 2017; Scaini et al. 2022).

Within the field of disaster risk reduction, “citizen seismology” (Bossu et al. 2011) constitutes a promising area, which is growing in different directions (Chen et al. 2020a). First, rapid information systems (e.g., smartphone apps, social networks) have proven to be reliable tools for collecting information when an earthquake is felt, such as its location, the degree of shaking, and damage. Regarding damage assessment, Sandron et al. (2021) present a method to rapidly draw maps of seismic impact on the basis of reports from trained volunteers in the Italian civil protection authority. Second, some projects have engaged citizens in the monitoring of seismic activity through the production of seismic data. For instance, Calais et al. (2020) developed a participatory seismology project based on easy-to-use seismological stations that are hosted by inhabitants of Haïti. Third, some projects have engaged citizens in the processing of seismic data, for instance, in the processing of P- and S- waves (Chen et al. 2020b). Participatory seismology has also been shown to foster the demand for information by the population and to spark interest in risk management (Calais et al. 2020).

Objectives and research questions

Mayotte is a French overseas territory (“département”) situated north of the Mozambique Channel and is characterized by a relatively low level of economic development, with more than 70% of the 256,500 inhabitants living below the poverty line as of 2017 (INSEE 2021). In May 2018, an unusual seismic crisis began, linked to the birth of a new submarine volcano (Cesca et al. 2020; Lemoine et al. 2020; Feuillet et al. 2021). This seismic activity was felt by the population for several months, and the new volcano is still active today. These events generated a great deal of anxiety in the population (Devès et al. 2021; Mori 2022).

The general objective of this study was to evaluate the potential for developing and sustaining a citizen science seismology project in Mayotte. In the long term, such a project might constitute a fruitful way to communicate with the public about the seismo-volcanic crisis, to foster dialogue between citizens and scientists, as well as being a relevant tool for engaging the local population in the production of scientific knowledge regarding Mayotte’s geophysical dynamics. In this paper, we explore both its technical feasibility (i.e., the reliability of the scientific protocol) and the conditions of its success in engaging Mayotte’s citizens (i.e., its perceived relevance and interest to the local population).

In summary, this citizen science project aims to engage the inhabitants of Mayotte in the manual identification of events recorded above the active volcanic system of Mayotte. Participants were invited to identify seismic events from a continuous time series of data collected using ocean bottom seismometers (OBS).

Our study was organized around two general research questions:

(Q1) Is the designed protocol adapted to produce reliable data by non-professional citizens, suitable for use in a citizen science project targeted at the general population?

(Q2) To what extent, and under which conditions, is this project able to reach different segments of Mayotte’s population?

It is worth noting that the answers to these research questions will also be of interest for the development of citizen seismology programs in other locations with similar contexts.

Socio-demographic context

Compared with metropolitan France, Mayotte is characterized by a relatively low level of economic development: 70% of the population live below the poverty line, and 25% have no formal settlements as of 2017 (INSEE 2021). Moreover, 71% of the population have no qualifying diploma, 53% are illiterate, and only 63% of people older than 14 years were French speakers in 2007 (Fallou et al. 2020); the most common local language is a Sabaki Bantu language, the Shimaore (Mori 2022). In addition, the socio-demographic situation in Mayotte is made more complex by the large diversity (and disparity) of its population (Walker 2019). First, Mayotte is located at a migratory crossroads with neighboring countries (Madagascar, Union of the Comoros) that are characterized by a much lower level of economic development and social protection (Roinsard 2014). Consequently, it is estimated that 40% of Mayotte’s inhabitants are foreigners (mostly from the Comoros), and 80% of them are there illegally (Roinsard 2014). Second, the local populations are subject to strong inequities, with the emergence of an educated middle class that has developed with the rise of public jobs since the island changed its status in 2011 and became a French département (Roinsard 2014).

Geophysical context

Before 2018, Mayotte was considered to be an area with moderate seismic activity, and the last earthquake was felt in 1993 (Bertil et al. 2021). In May 2018, intense and unexpected seismic activity started, with tens of earthquakes felt by the population that same month (BSCF 2018). The seismic crisis culminated with an earthquake of magnitude 5.9 on 15 May 2018. This seismicity was followed a month later by the start of a progressive subsidence of the island, associated with the deflation of a magma reservoir due to an eruption that started between June and July 2018 (Cesca et al. 2020; Lemoine et al. 2020; Feuillet et al., 2021). Then in May 2019, a new volcano, Fani Maoré, was discovered during a multi-institutional oceanographic campaign funded by the French government (Figure 1) (Feuillet et al. 2021). Since then, scientific missions have been regularly organized, both on land and offshore, to study the volcanic system dynamics where the activity is monitored by the REVOSIMA (Réseau de surveillance volcanologique et sismologique de Mayotte) (REVOSIMA 2021; REVOSIMA-IPGP 2021). From the newly recorded data, it was shown that the new volcano, Fani Maoré, is about 900 m high with a base at a depth of 3,000 m. The volcano is located 50 km southeast of Mayotte island. However, since 2018, seismic activity has been occurring offshore of Mayotte in two clusters, one around 10 km away and another one 25 km from Mayotte (Figure 1) (Saurel et al. 2022a). The seismic activity is monitored in real-time by OVPF (observatoire volcanologique du piton de la Fournaise) with support from the BRGM (French geological survey) in Mayotte (REVOSIMA-IPGP 2021). The seismicity is still ongoing today, with a daily average of 10 earthquakes, with about one every few months large enough to be felt. Due to some (limited) structural damage to buildings (BCSF 2018), these repeated earthquakes (seismic swarms) have resulted in anxiety and even panic events among a population that has not been accustomed to frequent earthquakes (Mori 2022).

Local perceptions of the seismo-volcanic crisis: state of the art

Mayotte’s local context is characterized by a complex interaction between scientific/safety information, and cultural/religion-related elements, as extensively shown by Cripps and Souffrin (2020), who reported on the perception of natural risks in Mayotte. Notably, these authors highlight the existence of traditional explanations for earthquakes (Leone 2014), and the conflicts or tensions among generations and social classes regarding the deference to scientific information. A few works have specifically studied the perceptions of the current seismo-volcanic crisis by Mayotte’s populations. Fallou et al. (2020) and Devès et al. (2021) focused on the feelings of inhabitants regarding scientific and safety information about the seismic swarms. Their results highlight the difficulty for the scientists and the authorities to reach the local population with information. Both studies argue that this lack of communication reinforces a widespread state of distrust or suspicion towards scientists and the authorities. These studies use the same empirical material: an online group of Mayotte’s citizens sharing information about perceived seismic activity (STTM group). Fallou et al. (2020) also circulated a survey among users of the LastQuake application, dedicated to earthquake information and the collection of reports of where earthquakes have been felt. However, these two samples exhibit specific biases. First, the STTM group may have an over-representation of inhabitants who have experienced a form of distrust or suspicion towards the authorities because of the self-selection process. Second, the LastQuake application might preferentially recruit citizens who complain about the lack of information. Hence, these studies are limited and should be completed by data collected through other channels in order to really study the local perceptions and perspectives regarding the seismic-volcanic crisis and its management. In particular, additional field work with Mayotte’s population is needed to assess the level of distrust and suspicion towards the scientific and administrative authorities, and the interest of citizens for stronger interactions with scientists through citizen science.

Seismicity analysis and design of the citizen science protocol

Seismicity analysis is one of the main tools used to study and monitor volcanic activity remotely. Indeed, seismicity is usually recorded before and during eruptions all over the world. In Mayotte, the seismicity is monitored daily at the OVPF observatory using an automatic detection process that analyzes the signals recorded by seismic stations installed on land (Retailleau et al. 2022a). The identification of earthquakes is checked and completed daily by the OVPF agent on duty, and their locations are validated by an analyst of the REVOSIMA earthquake location group.

Contrary to data from the land surface, which are available in almost real time, the data that are recorded from the ocean bottom by OBS stations are recovered every few months during oceanographic campaigns (REVOSIMA 2021). Automatic detection is applied afterwards to the data, but no manual screening is performed, apart from the larger earthquakes. These manual screening sessions performed by a group of seismologists have been called pickathons (Saurel et al. 2022a). Whereas land stations mostly record earthquakes, OBS stations record various types of events, including earthquakes, hydro-acoustic events, and other types of signals (Saurel et al. 2022a; Saurel et al. 2022b). Pickathons have also been organized to identify “exotic” events from OBS data over a period of a few weeks. Finally, the automatic detection that is being used to monitor the seismic activity has been shown to be very efficient for studying classical Volcano Tectonic earthquakes (VT), but it is more limited for studying another type of event that is linked to the volcanic activity: Long Period earthquakes (LP) (Retailleau et al. 2022b). For these reasons, manual screening is still crucial for understanding the ongoing seismicity in Mayotte and to validate the automatic analyses, but this is time consuming. Moreover, a significant part of the data has not yet been screened. Consequently, manual screening by citizens could be very useful for the advancement of scientific research in seismology in Mayotte, among other places.

Examples of the movement of the ground recorded by OBS stations are represented in Figure 2. To identify an event, one must determine the moment when the amplitude of the signal varies and emerges from the flat background level. To do this identification, it is not necessary to be an expert in this field. Hence, we expected that with some brief training, anyone should be able to do this task. The principle of our protocol is for the participants to screen continuous data and to identify events that have been recorded by several stations (the exact protocol circulated to the participants is in Supplemental File 1: Citizen science protocol). For this task, we use the WebObs platform, which permits users to scroll through hours of continuous signals (Beauducel et al. 2020). Figure 2 shows the platform and the seismic data. The main platform (Figure 2a) shows the aggregated data from all the stations displayed in 1-hour windows. Figure 2b shows an example from a one-hour window, and Figure 2c shows the data from a selected event.

In the citizen science protocol that we designed, each participant is assigned a sequence of continuous data (Figure 2), in which s/he can identify the seismic events. Each participant is provided with a computer and access to the OBS database through the WebObs platform. We wrote a French-English protocol to guide the participants, as one of the students was an English speaker (Supplemental File 1: Citizen science protocol). For our specific test, translation to Shimaore was not required.

Initial test with university students

The oceanographic campaign MAYOBS23 (Jorry et al. 2022) took place in July 2022. While these MAYOBS campaigns are recurrent and occur every few months, this one also hosted an “Ecole flottante” (on-board school), where students board the research vessel to learn about the work of scientists studying Mayotte’s volcanic activity. The 20 students (8 men and 12 women) were from Mayotte and the surrounding region, with the exception of a student from Paris. They were recruited to join the on-board school as part of a program independent from the authors of this study. Their study levels went from undergraduate to PhD. None of the students had any previous experience in seismology.

During this campaign, we tested our methodology to assess its clarity and efficiency. This kind of test is recommended prior to launching a citizen science protocol at a larger scale (Fraisl et al. 2022). We organized 4 sessions with an average of 5 students per session. For this first test, the participants analyzed seismic records from 19 April to 7 May 2019 (MAYOBS2; Jorry 2019).

We chose to study this timeframe because the data screening had not yet been done by scientists. Unfortunately, we did not have any data recorded by OBS stations from the beginning of the crisis in 2018. Each student was provided a laptop, and they identified events from different days during a period lasting 1 to 1.5 hours.

Semi-structured interviews

We conducted semi-structured interviews with 15 individuals (8 women, 7 men) who live in Mayotte to explore citizens’ perceptions of scientific and safety information and their motivations for participating in a citizen science seismology project. From this group, 13 interviews were done individually, and a single interview was performed with 2 persons simultaneously. Our sample consisted of the following 4 groups:

1. 3 students from Mayotte’s university who participated in the test of the citizen science protocol during MAYOBS23;
2. 5 high-level employees from public administrations—3 from municipalities, 1 from the local council, and 1 former employee from the Préfecture (local State government representation);
3. 6 presidents or employees of environmental or cultural associations; and
4. 1 journalist.

During the intense seismic swarms of 2018, 4 out of 15 of the participants (2 Mahorais people born on the Island and 2 individuals from Metropolitan France) were not present on the island. The interviewees were selected for their expertise regarding the sociological, cultural, and ethnographical features of Mayotte’s population owing to their central place in Mayotte’s social network, or their acquired expertise regarding the scientific protocol (for group a). Consequently, our sample does not aim to represent the average background of the target population. In addition, their relatively high level of education guaranteed easier communication in French. The recruitment of the interviewees was done by contacting local authorities, which provided us with a list of people (from municipal administrations and cultural associations). We designed two different interview guidelines. The first one was designed specifically for group a. It was composed of three blocks in which individuals were asked about (i) who they are and what they study; (ii) their memories of the seismic crisis; (iii) their feelings about the way that scientific and safety information was communicated to the population; and (iv) their feelings about the citizen science protocol. The second guideline was designed for the other groups (b, c, and d). These individuals were asked about (i) who they are and what their activities are in Mayotte; (ii) their perception of the political engagement of Mayotte’s citizens; (iii) their memories of the seismic crisis; (iv) their feelings about the way scientific and safety information was communicated to the population; and (v) their feelings about the opportunities to engage citizens in the citizen science seismology project that we are proposing.

All the interviews were started with a brief introduction to our study. The interviews lasted from 30 to 75 minutes. Their contents were transcribed and qualitatively assessed by thematic analysis (Nowell et al. 2017): Elements of the corpus were categorized into a certain number of themes and sub-themes, and we performed a discursive examination to draw an exhaustive map of topics and opinions expressed by the panel.

MAYOBS23 test

Results from the event identification sessions

We analyzed the event identification sessions made by the students during the oceanographic campaign. Figure 3a represents the number of events identified per student, which varied considerably. On average, each student worked on one to two hours of data, but this time period differed and could partly explain the variation in numbers. However, the seismic activity is also very varied, with periods of time that contain many events and others with none. Consequently, since each student analyzed a different window of time, there was no expectation that an identical number of events would be identified by each student. In total, 1,694 events were identified by the group during only four sessions, showing that citizen participation could greatly increase the dataset used by seismologists to study these events.

Figure 3 

a) Total number of events identified per participant. b) Total number of events identified per participant during the first hour in green. The dark green part of the bar shows the number of events added by an expert. Gray bars show participants for which this characterization could not be done.

To characterize the accuracy of identification for scientific use, we analyzed the quality of the identification by each student on the first hour of analysis. The aim was to assess the ability of the scientists to easily detect poor-quality contributions, that is, participants who identified only a small number of seismic events. Although these contributions are not scientifically unusable, the corresponding time series will need to be checked by experts to identify any missing events. Figure 3b represents the number of events per window identified by the students (light green) and added by the experts (dark green).

Each hour of data analyzed by the students could be checked by an expert in a few minutes. In two cases, the assignment was not properly understood, and although events were identified, the student did not work on the continuous hour but on the first portion of each hour of the day. Based on the rest of the group, 91.6% of events were flagged correctly, and 83.3% of the students identified more than 75% of events correctly. We note that there were no false identifications. Although some events were missed, this means that all the events identified were real earthquakes. Hence, all identified events could be used for scientific studies, and only little supplementary work was required by the scientists to produce the complete set of major events in the time series analyzed.

Perceptions of the seismo-volcanic crisis and potentials of our citizen seismology project

Table 1 presents a summary of the results. For every topic described in the text, we provide a short summary sentence, the number of times the topic occurred in the interviews, and some illustrative quotations from the interviewees.