The Amazon Biome in Messages --------------------------------- This project is developed by the digital innovations team of CGIAR Climate Security, Alliance of Bioversity International and CIAT. As part of the CGIAR Science Program on Climate Action, it aims to investigate how the Amazon Rainforest is discussed in public [Telegram channels (PT, EN, ES)][1] and the implications for public policy and information governance. ---------- **Data collection** - Telegram messages datasets were built using [TeleCatch][2], which extracts messages, metadata, and media files from public channels and groups. - Three distinct datasets were created by filtering the messages with the terms *amazon* and *amazonian* in English, in Portuguese, and in Spanish. ---------- **Initial data processing** - The first stage of analysis has been carried out with the dataset in Portuguese. - Initial data processing was conducted through [Telegram Analytics][3], an R-based reproducible pipeline that generates temporal trends, sentiment and emotion detection, influence and virality metrics, forwarding/mention/hashtag networks, topic modeling, and domain-sharing analysis. ---------- **Preliminary findings** The decision to begin with the Portuguese dataset is related to the centrality of Brazil in the global climate agenda. As the country that hosts the largest share of the Amazon Rainforest, Brazil plays a key role in shaping environmental policy and climate negotiations. This is particularly relevant in light of COP30, which will take place in Belém, in the Brazilian Amazon. By analyzing how the Amazon is discussed in Portuguese-language Telegram channels, the project provides insights into narratives, misinformation dynamics, and public communication practices that influence national debates and can significantly affect both domestic and international policy-making. Initial results provide a foundation for qualitative reading, channel classification, and policy recommendations on public communication and information integrity. The analysis comprises temporal trends, domain distribution, network analysis, sentiment analysis, and topic modeling. All result files and visualizations are available at [this folder][4], with the comprehensive [analytical report][5] accessible through the [Amazon in Messages – Portuguese Corpus Analysis][6] component of this project. ---------- **Future agenda** **1. Expand Multilingual Analysis** - Extend the methods already piloted with the Portuguese dataset to the English and Spanish corpora. - Compare narratives across languages to identify transnational disinformation flows and regional differences in how the Amazon is framed. **2. Deepen Network and Narrative Mapping** - Conduct network analyses of forwarding, mentions, and co-occurrence of domains/hashtags. - Map narrative clusters (e.g., sovereignty, conservation, denialism, development) and how they spread across different communities. **3. Integrate Qualitative Approaches** - Complement computational results with close readings of messages and channels. - Develop a typology of narratives and actors, connecting findings to policy frames. **4. Policy and Governance Applications** - Translate findings into policy briefs for decision-makers in Brazil and internationally. - Provide evidence-based recommendations for information integrity strategies in preparation for COP30 and beyond. **5. Disinformation Detection** - Develop and refine methods to detect climate- and Amazon-related disinformation in Telegram corpora. - Combine keyword filtering, machine learning classifiers, and OSINT-inspired approaches to identify recurring patterns, sources, and amplification strategies. - Build a reproducible framework that can be applied to other contexts of climate misinformation. **6. Open Science and Reproducibility** - Continue publishing datasets and analytical pipelines (e.g., via OSF + Zenodo), reinforcing open, transparent, and reproducible science. - Develop a methodological toolkit for other researchers investigating social and cultural issues on Telegram. ---------- **Acknowledgments** This project is conducted with support from the CGIAR Science Program on Climate Action and the CGIAR Science Program on Food Frontiers and Security. We would like to thank all funders who supported this research through their contributions to the [CGIAR Trust Fund][7]. [1]: https://osf.io/ys4j9/ [2]: https://github.com/labaffa/telecatch [3]: https://github.com/gtucci/Telegram-Analytics/tree/main [4]: https://osf.io/54suj/files/osfstorage [5]: https://osf.io/4cdfa [6]: https://osf.io/54suj/ [7]: https://www.cgiar.org/funders/