Study area

WildLIVE! is focused on data obtained from an ongoing camera trap biomonitoring project in the Chiquitano region, Bolivia, which was initiated in March 2017 (Jansen et al. 2020). Since its initiation, the monitored area was gradually expanded, and today comprises 25 camera trap stations, covering approximately 23,025 ha (Supplemental File 1: Figure 1). Each station was equipped with two (paired) cameras. This setup aimed to enhance the detection and identifiability of jaguars and other species based on individual coat patterns on each body side, with the jaguar as the focal species. For more details, please refer to former descriptions of the study site (Schulze et al. 2009) or the field methodology (Jansen et al. 2020) and analysis (Meißner et al. 2024).

WildLIVE! Development and construction

The WildLIVE! project is structured as follows (Supplemental File 2: Figure 2): Data is collected by camera traps and is subsequently pre-processed (e.g., long runs of empty images or images with humans present are removed) and then uploaded to the external web based platform Labelbox, a closed-source web-based labeling platform frequently used to create training data sets for machine learning applications (Labelbox 2021). Participants register for WildLIVE! through the landing page (see Jansen 2021), and are subsequently added to Labelbox. Labelbox granted us an extended free educational license for non-commercial research purposes. The platform allows for the development of a customizable interface for the manual classification of images, and subsequently, further inspection by other users. Upon registration, participants gain access to view the camera trap images online, accompanied by an integrated tutorial that was developed in collaboration with experienced participants (Jansen et al. 2021) to assist with species identification. Citizen scientists process images by assigning them to predefined categories or annotations, for example a species name, or as “empty” (referred to as “blanks” in other projects). If an image is found to contain one or more animal(s), a bounding box is drawn around each individual and is then assigned an identification (Figure 1). For the purpose of this paper, we use the term “label” for a classified image, in contrast to the various “annotations” that participants use to classify each image. For more details about “quality control” please refer to Supplemental File 3: Appendix 1.

To obtain an index of biodiversity, we used camera trap data collected between January 2017 and December 2021. For the consensus of image classifications that did not have a unanimous label classification, the image classifications made by “expert reviewers” were accepted, and for images not reviewed, classifications that reached a consensus of 70% or higher were accepted (Adam et al. 2021; Hsing et al. 2022). The resulting extensive data resource is being used for scientific analysis and as input to train an AI model for future classification aid.

Citizen scientists’ engagement and motivators

The research utilized a mixed-methods approach, collecting both quantitative and qualitative data to explore participants’ attitudes towards and involvement in WildLIVE!. To identify citizen scientist participation and engagement, we analyzed user data obtained from our online landing page and Labelbox. Participation was measured as the number of registrations over time between 5 April 2020 and 5 April 2023. Engagement was measured based on user data collected by Labelbox (“Created At,” “Seconds to Create,” “Seconds to Review,” “Created By;” for more details, refer to Supplemental File 3: Appendix 1).

Quantitative data were collected through a two-part questionnaire titled “Meine Teilnahme an WildLIVE!” (refer to Supplemental File 4: Appendix 2 for a more detailed description of its content and methodology) to understand the perceptions, motivations, and expectations of participants. The original version of the questionnaire was in German, but translated subsequently into English by two bilingual researchers in the subject matter who are fluent in both languages. This approach ensured that the nuances and contextual meanings of the questions were preserved (Bradley 1994).

The questionnaire was developed using a construct-centered approach (Messick 1995), ensuring that it accurately captured the constructs of interest. In the first section of the questionnaire, Part A, participants expressed their level of agreement with motivational statements crafted from ad-hoc email feedback received from citizen scientists. These statements were presented on a 6-point Likert scale ranging from 1 (do not relate) to 6 (strongly related; Zanbar and Ellison 2019). Motivational statements were carefully selected to be representative of the feedback received and to be aligned with the five constructs of knowledge and understanding, engagement and commitment, personal benefits and enjoyment, community and advocacy, and emotional response (Santori et al. 2021; Kim, Chan, and Gupta 2007; for examples of participants’ feedback via emails, see Supplemental File 5: Appendix 3).

Part B of the questionnaire offered open-ended questions to capture narrative responses, allowing for a richer understanding of participants’ experiences with WildLIVE! (see Supplemental File 3: Appendix 1 for more details). We carried out all analyses in R 3.6.1 (R Core Team, 2023) if not specified otherwise.

Citizen science participation, engagement, and the COVID-19 pandemic

The WildLIVE! initiative was launched on 4 April 2020 at the beginning of the COVID-19 pandemic as part of several other CS projects at the Senckenberg Institute for Biodiversity Research (Senckenberg 2023). Until 4 April 2023, 858 participants had registered. Active participants donated more than 8,938 hours of work, which equates to 1,117.3 working days or 4.98 years (assuming a 40-hour work week including German public holidays and vacation). In total, 108,198 images were processed with 10 to 11 classification iterations, resulting in 976,268 image classifications (labels) with an estimated 1,500,000 annotations assigned by citizen scientists.

The participants’ engagement had an intrinsic pattern affected by the national lockdowns during the COVID-19 pandemic (Supplemental File 6: Figure 3). Sudden spikes in registrations and engagement can be attributed to public outreach activities by us such as social media posts, print media articles, radio interviews, etc., as well as group registrations by university students during courses or school class registrations (Figure 2a). Bottlenecks in the upload of data in the initial phase caused a negative impact on citizen scientists’ engagement (Supplemental File 6: Figure 3).

Out of the 858 registered participants, 504 (59%) actively contributed to the classifications. The image classification–to–participant ratio was 1,937:1. The extent of the active citizen scientists’ contributions varied between 1 and 97,403 labels each (mean 1,943). Notably, a small number of participants were responsible for the majority of classifications: While 158 of the 504 active participants (31%) classified less than 100 images each, the most active 10% (50 individuals) contributed to 78% (n = 763,622) of all image classifications (labels). The top three participants alone accounted for an impressive number of labels, with individual throughputs of 97,403, 62,636, and 57,092 labels respectively (Figure 2b).

Analysis of taxonomic classifications

The taxonomic classifications for 42,088 images containing mammals were assigned by at least 10 or 11 different citizen scientists. Of these images, 29,437 (69.9%) were identified unanimously to species level. For 10,314 images (24.5%) with non-unanimous labels, the identification of expert reviewers was accepted. There were 1,437 images (3.4%) that were not reviewed by expert reviewers for which an identification was accepted if a consensus of 70% or higher was reached. 900 images (2.1%) that did not yield a consensus were discarded from this study.

Overall we recorded 24 mammalian species representing the native trophic structure including apex-predators (puma, jaguar), meso-carnivores (ocelots, jaguarundi, margays), large herbivores (tapir, collared and white lipped peccary), and small mammals (agouti, tamandua, nine-banded armadillo; Jansen et al. 2020). Because of the vulnerability of large carnivores to population decline and local extinction caused by habitat loss, we conducted a separate study (Meißner et al. 2024) focussing on the jaguar (Panthera onca). The study involved characterizing the life histories of 15 individuals and inferred four reproduction events between March 2017 to December 2019. The results of this separate study, which also determined a massive (33%) reduction of dense forest cover, provides strong evidence of the anthropogenic pressure on the CDF, and the imminent threat of habitat destruction in the Chiquitano region.

Citizen scientists’ motivations, incentives, and benefits

Out of 858 registered users, 504 were active participants, with 121 (24%) responding to the questionnaire. Participants predominantly expressed positive sentiments about their involvement in the WildLIVE! project, as reflected in the weighted average responses across construct themes. The constructs with the most positive responses were “Engagement and Commitment” (86.3% weighted average of positive responses for this construct), “Knowledge and Understanding” (83%), and “Emotional Responses” (78.4%). Responses in “Personal Benefits and Enjoyment” and “Community and Advocacy” were also positive, though to a lesser extent (68.7% and 55.6% respectively). The motivational statements with the three highest weighted averages of positive responses were (1) “gives me insights into wilderness” (92%), (2) “is a lot of fun” (91%), and (3) “is meaningful to me because I can support current research”, and “expands my knowledge on biological taxa/species” (89% for both, Figure 3; for presentation of responses presented on 6-point Likert scale, please refer to Supplemental File 7: Figure 4).The majority (85%) of participants expressed interest in continuing with the project, especially if new elements were introduced.

Conversely, there is also evidence that the project affects the participants in a negative way, with 83% feeling saddened by the observed biodiversity loss. Interestingly, 79% of participants had their first experience with CS through the WildLIVE! project. Yet, the project did not significantly improve awareness about other CS projects (25%) or contribute to reducing the feeling of wanderlust (35%; Figure 3).

In part B of the questionnaire, participants used a total of 3,127 words in the open-ended questions to describe their involvement, of which 467 were unique. Sentiment analysis revealed 69 positive, 31 negative, and 367 neutral words. The most frequently used words by participants to describe their experience were positive or neutral, including “knowledge” (75%), “interesting” (72%), “meaningful” (71%), “exciting” (63%) and “fascinating” (56%) (Table 1). Negative sentiments were reflected by the choice of words such as “monotonous” (26%), “exhausting” (23%), and “sad” (10%), as well as in personal communications conveying a decrease in motivation as a result of frustration about biodiversity loss (Figure 4; Supplemental File 8: Appendix D). These findings underscore the varied impact of WildLIVE!, from educational benefits to emotional challenges associated with conservation efforts (Table 1; Figure 3; see also Supplemental File 8: Appendix D).

Launching WildLIVE!

Virtual science projects have gained significant attention and utilization in biodiversity research, presenting an innovative and cost-effective approach to gathering and analyzing data. Participants are offered the opportunity to contribute to in-situ surveys, encompassing a variety of ecosystems and species, from freshwater habitats and avian populations to plant life (Rose et al. 2020; Dwivedi 2021; Ramvilas et al. 2021). They also facilitate ex-situ surveys in a digital space, where volunteers can assist, for example, in the classification of camera trap images and videos, such as “chimp&see,” “Snapshot Serengeti,” and “Wildlife Insights” (Swanson et al. 2015; Arandjelovic et al. 2016; Ahumada et al. 2020).

For most people, the COVID-19 pandemic demanded a shift from in-person nature observation to digital nature observation and may have changed how humans relate to the natural environment by driving us to observe nature from a more distant position (Poerting and Rudolfi 2021). A number of digital interactive projects such as BirdTrack (Shoesmith et al. 2021) and COVID-19 Sounds (Han et al. 2022) were launched as a result of the pandemic, and existing platforms such as eBird and iNaturalist (Kishimoto et al 2021; Sánchez-Clavijo et al. 2021) saw a marked increase in participation. In response to the COVID-19 pandemic, WildLIVE! was developed to offer opportunities for remote participation in biodiversity research, catering to the increased number of individuals seeking engagement from home during lockdown periods. In the context of this study, a “lockdown period” is defined based on the official restrictions imposed in Germany, where the study was conducted. These lockdown periods typically included government-mandated stay-at-home orders, closure of non-essential businesses, and restrictions on public gatherings.

It is important to note that while our analysis primarily draws upon the lockdown in Germany, we acknowledge the potential global reach of the WildLIVE! platform. Therefore, we do not assume that all participants are German or residing in Germany; however, the peaks in participant engagement from our data closely align with the lockdown periods in Germany, suggesting an impact of these restrictions on participation levels. Future studies could extend this analysis to examine the relationship between participant engagement and lockdown periods in other countries, providing a more comprehensive understanding of how global events and local policies influence digital CS participation.

While Labelbox provides a robust and reliable platform, it is important to highlight that it is a closed-source software. Therefore, risks are introduced such as vendor lock-in and limited interoperability, which might be significant considerations for some projects. Other platforms such as Zooniverse (www.zooniverse.org), eMammal (emammal.si.edu), and Agouti (agouti.eu) are open-source and potentially provide researchers with greater flexibility and control over their own projects. By opting for open-source solutions in the future, researchers can mitigate the risks associated with proprietary software platforms, fostering a more adaptable and resilient project environment.

At the moment, WildLIVE! is making use of the Labelbox platform to host the project until we have finished developing our own custom platform. This will be a combination of archive, analysis, and CS on one portal where data sets will be open source to enable open source research and CS according to the FAIR (Findability, Accessibility, Interoperability, and Reusability) principles (Wilkinson et al. 2016) and as described further in the section “Conclusion and Outlook.” One major benefit of the planned platform will be the open source availability of labeled camera trap datasets to train machine learning algorithms for automated classification.

Citizen science engagement and motivation

We found that the top 10% of participants (i.e., the 50 most active) contributed to 79% of image classifications (Figure 2b). This aligns with the outcomes of similar CS projects. For example, Sauermann and Franzoni (2015) analyzed participant engagement data across several projects and found that the majority of participants engaged only once, with minimal effort, whereas a small percentage of the participants contributed the greatest effort. The same trend was observed in the project MammalWeb (Hsing et al. 2022). However, in comparison with similar projects, WildLIVE! has a relatively high image-classification-to-participant ratio (~1,900:1) relative to MammalWeb (~550:1), and Snapshot Serengeti (~43:1), with 1.2 million images for 28,000 participants (Hsing et al. 2018; Swanson et al. 2016). On the one hand, this shows a huge potential for growth of WildLIVE! in terms of participant numbers in comparison with similar projects. On the other hand, it demonstrates a strong commitment resulting in heavy engagement of single participants. Approximately a year after the launch of WildLIVE!, the cumulative number of labels generated exceeded the number of participant registrations (Figure 2a), evidencing this strong engagement and activity by our participants.

Several studies have indicated the positive impacts of CS projects, including direct personal benefits and other broader outcomes such as behavioral alterations (see Green et al. 2020 for a review). Through involvement in CS projects, participants can increase their knowledge about the topics researched (Masters et al. 2016; Forrester et al. 2017), and provide a sense of purpose and community (Curtis 2015; Domroese and Johnson 2017). While we did not conduct an extensive or systematic survey on the underlying motivations of our participants, we received valuable personal feedback through our questionnaire, as well as through ad hoc email correspondence (see some examples in Supplemental File 5: Appendix 3). The results of our questionnaire indicate that the primary motivations of our participants, reflected by the statements “insights into wilderness, “expands my knowledge,” and “meaningful” align predominately with the categories of “Engagement and Commitment,” followed by “Knowledge and Understanding” and “Emotional Responses.” The words most commonly used by participants to describe their involvement in WildLIVE! include “knowledge,” “species,” “exciting,” and “interesting.” This reinforces the idea that understanding and knowledge, purposeful engagement, and enjoyment are key drivers motivating our participants, and suggests that our participants appear to be driven primarily by a desire for intellectual growth and emotional fulfillment.

Conversely, some participants are affected in a negative way, feeling saddened by the observed biodiversity loss. Involvement in the project has also been described as “monotonous” and “exhausting.” While their engagement in WildLIVE! represented the first interaction with CS for most participants, it did not improve their awareness about other CS projects or reduced the feeling of wanderlust. Our participants are becoming increasingly concerned about habitat destruction occurring in or nearby our study area, and occasionally reported a resulting decrease in motivation. Unfortunately, this is in line with other studies reporting on a growing body of evidence relating to emotional distress responses arising from perceptions on environmental change in society (Cianconi et al. 2022).

Attract new and retain existing participants

In our experience, the two major challenges in the WildLIVE! project are (1) to attract new participants and (2) to retain existing and experienced users. To draw new participants, social media platforms are crucial for the distribution of information about the project (Liberatore et al. 2018). To promote our project, our institution’s public relations department used social media platforms such as Facebook (facebook.com/SenckenbergWorld), Twitter (twitter.com/Senckenberg), and Instagram (instagram.com/SenckenbergWorld), as well as newsletters through the mailing list for members of the Senckenberg society. WildLIVE! also has its own webpage (Jansen 2021), which is linked to several public portals for CS (Bürger Schaffen Wissen; engagement jetzt!; GoVolunteer). Important but somewhat ad hoc is the outreach gained through mainstream media coverage, for example, print media or radio interviews. These spontaneous mainstream media exposure events are usually followed by a sudden increase in participant registrations (Figure 2a).

To enhance the retention and engagement of participants already registered in the WildLIVE! project, we implemented a variety of strategies, which have demonstrated promising results. Below, we describe these strategies along with practical examples of how they were applied:

1. Positive introduction: Providing a welcoming tutorial and introductory email helps create a positive first impression, setting the tone for participant involvement.
2. Personal care: Responding personally to emails, inquiries, and telephone calls from participants demonstrates our commitment to their positive experience, fostering a supportive community.
3. Dynamic and enriching experience: Facilitating digital access allows for time-flexible engagement, while providing content related to biodiversity and conservation keeps the experience relevant and educational. For example, we put effort into creating a dynamic home page, including news about the study area and an online museum of selected photographs.
4. Ownership and purpose: Encouraging co-design of project aspects, such as writing specific aspects of the tutorial (many participants intrinsically offered to write species identification keys), and participation in small sub-projects provides participants with a sense of ownership and purpose. Participants can become “expert reviewers” by classifying > 3 000 images within a project, providing acknowledgment of their efforts and additional motivation.
5. Clear impact goals: Communicating research questions, preliminary results, and conservation issues in an accessible and authentic manner ensures that participants understand the impact and importance of their contributions.
6. Transparency: Maintaining transparency in project development and providing frequent updates through newsletters (~ twice a year) and email correspondence have been instrumental in building trust and maintaining participant interest.
7. Appreciation and feedback: Showing appreciation for participant involvement through special acknowledgments, small gifts (e.g., a plaster cast of a jaguar track), and providing feedback on their contributions helps validate their efforts and encourages continued participation.
8. Community and exchange: Hosting online events and regular online meetings between project management and participants promotes a sense of community, facilitates knowledge exchange, and provides opportunities for co-design. Additionally, WildLIVE! has been successfully integrated into school enrichment events covering a range of curriculum subjects such as geography, biology, and ecology.
9. Visability, range, and outreach: WildLIVE! is hosted and promoted by Senckenberg – Leibniz Institution for Biodiversity and Earth System Research. The project also appears on two CS platforms, with the national platform BürgerSchaffenWissen attracting a large and growing user base. The project’s outreach, detailed in Figure 2, has positively impacted the attraction of new participants. In addition, a public and coached workspace in the “Aha?! Science Lab” hands-on exhibition at the Senckenberg Natural History Museum in Frankfurt, Germany, along with school events, continuously recruits new participants (Supplemental File 9: Figure 5).
10. Sustainability through institutional support and commitment: Embedded in Senckenberg’s broader CS initiative (“Gemeinsam forschen”), WildLIVE! along with various other CS projects benefit from coordinated support and a dedicated press office. The project is also showcased in the public museum (Supplemental File 9: Figure 5A), highlighting institutional commitment and support.

The effectiveness of these strategies is reflected in the observed peaks in participant engagement, the positive feedback received, as well the relatively high engagement (e.g., the high image-per-labeler ratio) and time investment of the participants, indicating a successful implementation of engagement-enhancing practices in the WildLIVE! project. A wealth of studies suggest that employing methods similar to those previously described can cultivate emotional connections with participants, thereby substantially boosting engagement in CS initiatives (Baruch et al. 2016; Hsing et al. 2022; Hart et al. 2022).

Citizen scientists explore “digital ecology”

The WildLIVE! project fosters collaboration between professional and citizen scientists in a digital space, driven by a shared interest in biodiversity and conservation. Citizen scientists enjoy the exposure to a diversity of exotic animals, and in particular the jaguar, a focal species of the project, which is not only of ecological importance but also boasts aesthetic charisma for people (Lorimer 2007), making the species particularly suitable for CS projects.

The CDF is a topographical space in which we conduct research on jaguars and other wildlife. However, WildLIVE! provides an additional layer as a topological space where characteristics of medialized human-animal encounters allow for further analysis of engagement (Pütz and Schlottmann 2020). In this context, the jaguar is not only significant as an important variable in the ecosystem, but also as an “object” that brings different working groups across differing professional backgrounds together in digital space for heterogeneous cooperation.

While our initial results have a primary focus on jaguar biology and population dynamics (Jansen et al. 2020; Meißner et al. 2024), the jaguar may also be regarded as a flagship species that functions as a boundary object, “where each social world has partial jurisdiction over the resources represented by that object” (Star and Griesemer 1989, p. 412). This perspective opens the discussion towards more socio-ecological research, and takes into account the social dimensions of CS. It underlines the importance of integrating societal actors, such as landowners and farmers. Since our analysis shows the need to take measures to prevent the further deforestation of CDF, this issue remains open on a sociopolitical level.

The WildLIVE! project, while not directly connected to tangible biodiversity conservation efforts, plays an important role in promoting biodiversity conservation by raising awareness. For example, citizen scientists might reflect on their environmental impact and consumption habits. Additionally, it provides valuable insights for decision-makers like landowners and farmers by providing evidence of the extent and impacts of land use changes on wildlife. Participant involvement in WildLIVE! plays a critical role in connecting individuals from urban environments with wildlife in a real “hands-on” way. Thus, participant engagement in this project promotes a connectedness with nature and builds a sense of responsibility towards our natural heritage. And on the flip side, citizen scientists are playing a critical role in enabling research by aiding the processing of a large data set. We need the continued participation and engagement of citizen scientists for the sustainability of long-term biomonitoring that can be used to track the effects of land use change in the CDF and possibly other areas.