Elamur Research⌕
☰
home / papers / 2026 / How Loud Rumbles Hit Newsstands: A Data Analysis of Coverage and Spatial Bias in German News about Landslides Around the World

How Loud Rumbles Hit Newsstands: A Data Analysis of Coverage and Spatial Bias in German News about Landslides Around the World

Paper · 2026 · Brielen Madureira, Andreas Niekler, Marc Keuschnigg et al.

Summary

The paper asks how German newspapers cover landslides outside Germany, and whether that coverage matches how often such disasters happen. The authors start with a wiso-net keyword retrieval of German articles from 2000–2024. They filter it with a locally run open-weight LLM (google/gemma-4-31b) in two stages: first deciding whether a text reports real landslides, then extracting the affected countries as ISO-3 codes. Both stages are checked against a human-annotated sample.

Each country's daily article counts are then split into "news events" with a simple gap rule. The event counts are compared with two external references: the EM-DAT disaster inventory and the World Bank's Global Landslide Hazard Map. The final dataset has roughly 55k articles grouped into about 4.5k news events. Coverage turns out to be uneven: Southern and Western Europe get more attention than their landslide occurrence would predict, while other regions are overlooked.

This matters in two ways. News-derived data can fill gaps in incomplete global disaster databases. It also gives a measurable basis for studying inequality in media attention to international disasters.

Key points

  • The pipeline starts from 317,036 retrieved articles (17 German landslide keywords). Preprocessing leaves 88,940 candidates, and 450 of them were human-annotated (58.6% relevant).
  • On the gold sample, relevance classification reached precision 0.865, recall 0.973 and F1 0.916. Country geolocation reached 0.875 for exact matches and 0.924 when any country overlapped.
  • A news event is a run of active days in one country's daily count series. Five or more consecutive days without articles start a new event, so gaps of up to four days are tolerated within one event.
  • The external comparison covers 244 countries. Indicators include EM-DAT landslide counts (2,071 events for 2000–2024), WB-GLHM predicted annual frequency, ThinkHazard! risk class, UNCTAD development status and World Bank income group.
  • The texts cannot be shared and must be accessed through wiso-net. The code is published on Codeberg (briemadu/landslides-german-news).

Interesting ideas

  • I1. The LLM is not asked for everything at once: it first gives a cheap yes/no relevance decision, and only accepted texts go to a second prompt that returns strict JSON. This keeps the extraction step focused and each stage can be scored separately.
  • I2. The authors read all 40 false positives by hand and found that most were annotation mistakes or texts about landslide risk, not real model errors. Auditing disagreements before trusting a metric shows whether the gold standard or the model is at fault.
  • I3. To show that a lenient "any overlap" accuracy was not inflated by long lists of guessed countries, they compared how many countries per document the model output versus the gold labels. This acts as a sanity control on the metric itself.
  • I4. Raw mentions are grouped into episodes by a single gap threshold, and the threshold is stated openly as a hyperparameter with clear trade-offs (too low splits one story, too high merges separate ones).
  • I5. The results are validated against two references of different kinds: an inventory of observed, impact-filtered events and a predictive environmental model. Disagreements between them, such as Canada, are explained from how each source was built.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (learn): LLM teacher tiers could be run the same way, with a gate question first and a structured silver label only for items that pass, each tier with its own score on gold.
  • I2 (learn): when witness parsers disagree with a treebank, manually triaging a sample would separate parser faults from gold errors before rules or training data are changed.
  • I3 (answer): QA scoring that accepts partial matches could include a check that the answer-set sizes match gold, so over-listing cannot game the score.
  • I4 (world): mentions of the same happening scattered through a long text could be clustered into one event by a transparent distance threshold, keeping event ordering explainable.
  • I5 (world): facts compiled from Wikidata and WordNet/FrameNet could be cross-checked against each other, with conflicts traced to each source's inclusion criteria rather than silently resolved.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Leipzig UniversityLeipzig · Germany26 works · 8,622 citations🎓Linköping UniversityLinköping · Sweden18 works · 2,319 citations🔬Helmholtz Centre for Environmental ResearchLeipzig · Germany3 works · 1 citation

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Social Sciences · 1,772 worksCommunication · 97 worksEnvironmental Science · 9 worksManagement, Monitoring, Policy and Law · 2 works

Cite

Cite as Brielen Madureira, Andreas Niekler, Marc Keuschnigg and Mariana Madruga de Brito. 2026. How Loud Rumbles Hit Newsstands: A Data Analysis of Coverage and Spatial Bias in German News about Landslides Around the World. arXiv:2605.18105.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2605.18105 · arXiv 2605.18105

@misc{madureira2026how,
  title = {{How Loud Rumbles Hit Newsstands: A Data Analysis of Coverage and Spatial Bias in German News about Landslides Around the World}},
  author = {Brielen Madureira, Andreas Niekler, Marc Keuschnigg and Mariana Madruga {de Brito}},
  year = {2026},
  doi = {10.48550/arxiv.2605.18105},
  eprint = {2605.18105},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2605.18105}
}

References

46 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Pedro Henrique Lima Alencar, Jan Sodoge, Eva Nora Paton, and Mariana Madruga De Brito. 2024. Flash droughts and their impacts—using newspaper articles to assess the perceived consequences of rapidly emerging droughts. Environmental Research Letters, 19(7):074048. ↗
  2. Shayan Alipour, Niccolò Di Marco, Michele Avalle, Gabriele Etta, Matteo Cinelli, and Walter Quattrociocchi. 2024. The drivers of global news spreading patterns. Scientific Reports, 14(1). ↗
  3. Aydoğan Avcıoğlu, Ogün Demir, and Tolga Görüm. 2025. An automated approach for developing geohazard inventories using news: integrating natural language processing (nlp), machine learning, and mapping. Natural Hazards and Earth System Sciences, 25(7):2421–2435. ↗
  4. Alessandro Battistini, Ascanio Rosi, Samuele Segoni, Daniela Lagomarsino, Filippo Catani, and Nicola Casagli. 2017. Validation of landslide hazard models using a semantic engine on online news. Applied Geography, 82:59–65. ↗
  5. Tal Bauman, Yuval Fisher, Fadi Kizel, Itzhak Omer, and Sagi Dalyot. 2025. Geographic information retrieval: a systematic review of datasets and their impact on textual geolocation. Geo-spatial Information Science, page 1–18. ↗
  6. José Eduardo Bonini, Tiago Damas Martins, Marina Tamaki de Oliveira Sugiyama, and Bianca Carvalho Vieira. 2025. Landslide inventory of the 2023 serra do mar disaster (brazil). Discover Geoscience, 3(1). ↗
  7. Erica Cai, Xi Chen, Reagan Grey Keeney, Ethan Zuckerman, Brendan O’Connor, and Przemyslaw A. Grabowicz. 2025. Identifying and investigating global news coverage of critical events such as disasters and terrorist attacks. Proceedings of the International AAAI Conference on Web and Social Media, 19(1):307–323. find
  8. Richard A. Davis, Konstantinos Fokianos, Scott H. Holan, Harry Joe, James Livsey, Robert Lund, Vladas Pipiras, and Nalini Ravishanker. 2021. Count time series: A methodological review. Journal of the American Statistical Association, 116(535):1533–1547. ↗
  9. Damien Delforge, Valentin Wathelet, Regina Below, Cinzia Lanfredi Sofia, Margo Tonnelier, Joris A.F. van Loenhout, and Niko Speybroeck. 2025. Em-dat: the emergency events database. International Journal of Disaster Risk Reduction, 124:105509. ↗
  10. María José Domínguez Cuesta, Montserrat Jiménez Sánchez, and Augusto Rodríguez García. 1999. Press archives as temporal records of landslides in the north of spain: relationships between rainfall and instability slope events. Geomorphology, 30(1):125–132. ↗
  11. Harideva Marturano Egas, Rodrigo Augusto Stabile, Marcio Roberto Magalhães de Andrade, Gean Paulo Michel, João Paulo Carvalho de Araújo, Rossano Dalla Lana Michel, Tatiana Sussel Gonçalves Mendes, Tulius Dias Nery, Danielle Silva de Paula, and Elisabete Weber Reckziegel. 2024. Comprehensive inventory and initial assessment of landslides triggered by autumn 2024 rainfall in rio grande do sul, brazil. Landslides, 22(2):579–589. ↗
  12. Robert Emberson, Dalia Kirschbaum, and Thomas Stanley. 2020. New global characterisation of landslide exposure. Natural Hazards and Earth System Sciences, 20(12):3413–3424. ↗
  13. Rachele Franceschini, Ascanio Rosi, Filippo Catani, and Nicola Casagli. 2022. Exploring a landslide inventory created by automated web data mining: the case of italy. Landslides, 19(4):841–853. ↗
  14. Melanie J. Froude and David N. Petley. 2018. Global fatal landslide occurrence from 2004 to 2016. Natural Hazards and Earth System Sciences, 18(8):2161– 2181. ↗
  15. Melanie Gall, Kevin A. Borden, and Susan L. Cutter. 2009. When do losses count?: Six fallacies of natural hazards loss data. Bulletin of the American Meteorological Society, 90(6):799 – 810. find
  16. Johan Galtung and Mari Holmboe Ruge. 1965. The structure of foreign news. Journal of Peace Research, 2(1):64–90. ↗
  17. Stefano Luigi Gariano and Fausto Guzzetti. 2022. Mass-Movements and Climate Change, page 546–558. Elsevier. find
  18. Derly Gómez, Edwin F. García, and Edier Aristizábal. 2023. Spatial and temporal landslide distributions using global and open landslide databases. Natural Hazards, 117(1):25–55. ↗
  19. Lynn M Highland and Peter Bobrowsky. 2008. The landslide handbook - A guide to understanding landslides. Circular 1325. U.S. Geological Survey. ↗
  20. Ruihong Huang, Ignacio Cases, Dan Jurafsky, Cleo Condoravdi, and Ellen Riloff. 2016. Distinguishing past, on-going, and future events: The EventStatus corpus. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 44–54, Austin, Texas. ACL. find
  21. Rebecca Louise Jones, Aditi Kharb, and Sandy Tubeuf. 2023. The untold story of missing data in disaster research: a systematic review of the empirical literature utilising the emergency events database (em-dat). Environmental Research Letters, 18(10):103006. ↗
  22. Stijn Joye. 2010. On the media construction of international disasters. Tijdschrift voor Communicatiewetenschap, 38(2):139–155. find
  23. Dalia Bach Kirschbaum, Robert Adler, Yang Hong, Stephanie Hill, and Arthur Lerner-Lam. 2009. A global landslide catalog for hazard applications: method, results, and limitations. Natural Hazards, 52(3):561–575. ↗
  24. Oddvar Kjekstad and Lynn Highland. 2009. Economic and Social Impacts of Landslides, page 573–587. Springer Berlin Heidelberg. find
  25. Inhye Kong and Ross S. Purves. 2026. Analyzing geographic bias of newspaper articles reporting global climate disasters. Annals of the American Association of Geographers, 116(2):270–288. ↗
  26. Thomas M. Kreuzer and Bodo Damm. 2020. Automated digital data acquisition for landslide inventories. Landslides, 17(9):2205–2215. ↗
  27. Lukas Kriesch and Sebastian Losacker. 2025. A geolocated dataset of German news articles. Scientific Data, 12(1):1128. ↗
  28. Ni Li, Shorouq Zahra, Mariana Brito, Clare Flynn, Olof Görnerup, Koffi Worou, Murathan Kurfali, Chanjuan Meng, Wim Thiery, Jakob Zscheischler, Gabriele Messori, and Joakim Nivre. 2024. Using LLMs to build a database of climate extreme impacts. In Proceedings of the 1st Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2024), pages 93–110, Bangkok, Thailand. ACL. find
  29. Brielen Madureira, Mariana Madruga de Brito, and Andreas Niekler. 2026a. Retrieving floods without floodlights: Topic models as binary classifiers for extreme climate events in German news. In Proceedings of the 2nd Workshop on Ecology, Environment, and Natural Language Processing, pages 115–134, Palma de Mallorca, Spain. European Language Resources Association. ↗
  30. Brielen Madureira, Andreas Niekler, Marc Keuschnigg, and Mariana Madruga de Brito. 2026b. The newsworthiness of brazilian distress: A peak analysis on time series of international media attention to disasters in brazil. Preprint, arXiv:2605.04552. ↗
  31. Tapas Ranjan Martha, Priyom Roy, Nirmala Jain, Kirti Khanna, K. Mrinalni, K. Vinod Kumar, and P. V. N. Rao. 2021. Geospatial landslide inventory of india—an insight into occurrence and exposure on a national scale. Landslides, 18(6):2125–2141. ↗
  32. Chris Massey, Kerry Leith, Tom R Robinson, Biljana Lukovic, Sam McColl, Trevor Carey-Smith, Brenda Rosser, Liam Wotherspoon, Hugh Smith, Harley Betts, Rob Buxton, and Janine Bidmead. 2025. What controlled the occurrence of more than 116, 000 human-mapped landslides triggered by cyclone gabrielle, new zealand? Landslides, 22(12):3953–3972. ↗
  33. Benjamin B. Mirus, Eric S. Jones, Rex L. Baum, Jonathan W. Godt, Stephen Slaughter, Matthew M. Crawford, Jeremy Lancaster, Thomas Stanley, Dalia B. Kirschbaum, William J. Burns, Robert G. Schmitt, Kassandra O. Lindsey, and Kevin M. Mc-Coy. 2020. Landslides across the usa: occurrence, susceptibility, and data limitations. Landslides, 17(10):2271–2285. ↗
  34. Kieu Anh Nguyen, Yi-Jia Jiang, Chiao-Shin Huang, Meng-Hsun Kuo, and Walter Chen. 2024. Leveraging internet news-based data for rockfall hazard susceptibility assessment on highways. Future Internet, 16(8):299. ↗
  35. Priscilla Niyokwiringirwa, Luigi Lombardo, Olivier Dewitte, Axel A. J. Deijns, Nan Wang, Cees J. Van Westen, and Hakan Tanyas. 2024. Event-based rainfall-induced landslide inventories and rainfall thresholds for malawi. Landslides, 21(6):1403–1424. ↗
  36. Einar Östgaard. 1965. Factors influencing the flow of news. Journal of Peace Research, 2(1):39–63. ↗
  37. David Petley. 2012. Global patterns of loss of life from landslides. Geology, 40(10):927–930. ↗
  38. Karl Erik Rosengren. 1970. International news: Intra and extra media data. Acta Sociologica, 13(2):96–109. ↗
  39. Denise J. Roth, Nan Bai, Robbert Biesbroek, Art Dewulf, Sanne Kruikemeier, Daan de Leur, Mariken A. C. G. van der Velden, Erik de Vries, and Rens Vliegenthart. 2025. Unveiling climate adaptation: A comprehensive analysis of agenda-setting dynamics in the united kingdom and the netherlands. Science Communication. ↗
  40. Jan Sodoge, Christian Kuhlicke, and Mariana Madruga De Brito. 2023. Automatized spatio-temporal detection of drought impacts from newspaper articles using natural language processing and machine learning. Weather and Climate Extremes, 41:100574. ↗
  41. Faith E. Taylor, Bruce D. Malamud, Katy Freeborough, and David Demeritt. 2015. Enriching great britain’s national landslide database by searching newspaper archives. Geomorphology, 249:52–68. ↗
  42. Bram Valkenborg, Olivier Dewitte, and Benoît Smets. 2026. Unravelling information on impactful geo-hydrological hazard events with hazminer, a multilingual text mining method developed through a global scale coverage application. EGUsphere, 2026:1–45. ↗
  43. Miet Van Den Eeckhaut, Javier Hervás, and Luca Mon-tanarella. 2013. Landslide Databases in Europe: Analysis and Recommendations for Interoperability and Harmonisation, page 35–42. Springer Berlin Heidelberg. find
  44. Chenchen Xie, Yuandong Huang, Chong Xu, Kebin Dai, and Xiwei Xu. 2026. Over 100, 000 landslides triggered by typhoon-induced rainfall in north china in july 2023. Landslides, 23(5):1389–1408. ↗
  45. M. M. Yagoub, Aishah A. Alsereidi, Elfadil A. Mohamed, Punitha Periyasamy, Reem Alameri, Salama Aldarmaki, and Yaqein Alhashmi. 2020. Newspapers as a validation proxy for gis modeling in fujairah, united arab emirates: identifying flood-prone areas. Natural Hazards, 104(1):111–141. ↗
  46. Yan Yan and Kim Bissell. 2015. The sky is falling: Predictors of news coverage of natural disasters worldwide. Communication Research, 45(6):862–886. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/madureira-2026-loud-rumbles-hit-newsstands-data/