Abstract
Traditional natural disaster response involves significant coordinated teamwork, where speed and efficiency are key. Nonetheless, human limitations can delay critical actions and inadvertently increase human and economic losses. Agentic Large Vision Language Models (LVLMs) offer an avenue to address this challenge, with the potential for substantial socio-economic impact, particularly by improving resilience and resource access in underdeveloped regions. We introduce DisasTeller, a multi-LVLM-powered framework designed to automate tasks in post-disaster management, including on-site assessment, emergency alerts, resource allocation, and recovery planning. By coordinating four specialised LVLM agents with GPT-4 as the core, DisasTeller can accelerate disaster response activities, reducing human execution time and structuring information flow. Our evaluation shows both benefits and challenges: while DisasTeller streamlines coordination and report generation, errors in early-stage assessments may propagate downstream, highlighting the need for human validation and improved LVLM accuracy. This framework acts as a complementary support system to expert teams, bridging the gap between traditional response methods and emerging LVLM-driven efficiency, while highlighting the importance of continued refinement and collaboration for safe, trustworthy deployment.
Similar content being viewed by others
Data availability
The crisis image dataset in the main text is released by Qatar Computing Research Institute for humanitarian computing research20, and can be accessed from the official project website: https://crisisnlp.qcri.org/data/ASONAM17_damage_images/ASONAM17_Damage_Image_Dataset.tar.gz. The bushfire and flood images in Supplementary Information are released by the Australian government report41,42: https://flooddata.ses.nsw.gov.au/flood-projects/post-2022-event-flood-behaviour-analysis-tweed-river-report-only, https://www.fire.qld.gov.au/sites/default/files/2021-04/Response-Magazine-Dec2019_0.pdf. The license terms for the crisis image dataset in the main text, as well as the flood and bushfire images in the Supplementary Information, can be accessed at the following links: https://crisisnlp.qcri.org/terms-of-use.html, https://flooddata.ses.nsw.gov.au/related-dataset/post-2022-event-flood-behaviour-analysis-tweed-river-report/resource/4faca39c-2163-4dde-a6b5-340229e9ae6b by a pop-up window during download, and https://www.fire.qld.gov.au/copyright. According to the licences, the raw data can be used for research purposes but cannot be publicly shared. Therefore, all image samples shown in this paper are synthetic illustrations and not taken from original datasets. The output data of DisasTeller are available at the Zenodo repository: https://doi.org/10.5281/zenodo.17785783. Source data are provided with this paper.
Code availability
All code and synthetic data supporting this study are available in the GitHub repository: https://github.com/zche3016/DisasTeller/tree/main.
References
Uncounted costs: data gaps hide the true human impacts of disasters in 2023, United Nations Office for Disaster Risk Reduction. https://www.undrr.org/explainer/uncounted-costs-of-disasters (2023).
Sendai Framework for Disaster Risk Reduction 2015-2030, United Nations Office for Disaster Risk Reduction. https://www.undrr.org/publication/sendai-framework-disaster-risk-reduction-2015-2030 (2015).
Natural Disaster and Emergency Learnings and Recommendations Report, New South Wales Rural Doctor Network, Australia. https://www.nswrdn.com.au/client_images/2244880.pdf (2021).
Edwin, M. L., Naomi, K. City and County of San Francisco Post Disaster Safety Assessment Guide. https://www.sf.gov/sites/default/files/2022-06/Post%20Disaster%20Safety%20Assessment%20Guide%202.0.June2016.pdf (2016).
Ao, Y., Bahmani, H. Current Research Status of Disasters and Human Response, Navigating Complexity: Understanding Human Responses to Multifaceted Disasters, pp 1–18 (2023).
Feng, Y. & Cui, S. A review of emergency response in disasters: present and future perspectives. Nat. Hazards 105, 1109–1138 (2021).
Park, S., Yao, T. & Ukkusuri, S. V. Spatiotemporal heterogeneity reveals urban-rural differences in post-disaster recovery. npj Urban Sustain. 4, 2 (2024).
Forrester, N. Meet the scientists planning for disasters. Nature 619, S1 (2023). https://www.nature.com/articles/d41586-023-02312-2.
Pollack, A. B., Helgeson, C., Kousky, C. & Keller, K. Developing more useful equity measurements for flood-risk management. Nat. Sustain. 7, 823–832 (2024).
Ceferino, L., Merino, Y., Pizarro, S., Moya, L. & Ozturk, B. Placing engineering in the earthquake response and the survival chain. Nat. Commun. 15, 4298 (2024).
Opabola, E. A. & Galasso, C. Informing disaster-risk management policies for education infrastructure using scenario-based recovery analyses. Nat. Commun. 15, 325 (2024).
Ding, M. et al. Reversal of the levee effect towards sustainable floodplain management. Nat. Sustain. 6, 1578–1586 (2023).
Achiam, J. et al. Gpt-4 technical report. Preprint at arXiv https://arxiv.org/abs/2303.08774 (2023).
Bran, A. M. et al. Augmenting large language models with chemistry tools. Nat. Mach. Intell. 6, 525–535 (2024).
Bharathi Mohan, G. et al. An analysis of large language models: their impact and potential applications. Knowl. Inf. Syst. 66, 5047–5070 (2024).
Brown, T. et al. Language models are few-shot learners. Adv. Neural Inf. Process. Syst. 33, 1877–1901 (2020).
Radford, A. et al. Language models are unsupervised multitask learners. OpenAI Blog 1, 9 (2019).
Liu, Y. et al. G-Eval: NLG Evaluation using gpt-4 with better human alignment. In Proc. 2023 Conference on Empirical Methods in Natural Language Processing, pp 2511–2522 (2023).
Yang, J. et al. Poisoning medical knowledge using large language models. Nat. Mach. Intell. 6, 1156–1168 (2024).
Nguyen, D. T., Ofli, F., Imran, M., Mitra, P. Damage assessment from social media imagery data during disasters. Proceedings of the 2017 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, pp 569–576 (2017).
Grünthal, G. European Macroseismic Scale 1998 (EMS-98). https://www.gfz-potsdam.de/en/section/seismic-hazard-and-risk-dynamics/data-products-services/ems-98-european-macroseismic-scale (1998).
Bashawri, A., Garrity, S. & Moodley, K. An overview of the design of disaster relief shelters. Proc. Econ. Financ. 18, 924–931 (2014).
FEMA, Historic Disasters, Federal Emergency Management Agency. https://www.fema.gov/disaster/historic (2025).
Chen, L., Zaharia, M. Zou, J. How is ChatGPT’s behavior changing over time? Harvard Data Sci. Rev. https://arxiv.org/abs/2307.09009 (2024).
Team, G. et al. Gemma 3 Technical Report. https://arxiv.org/abs/2503.19786 (2025).
Ninivaggi, G. Ishikawa earthquake response sees growing opposition scrutiny. The Japan Times https://www.japantimes.co.jp/news/2024/01/11/japan/politics/kishida-disaster-response/ (2024).
Augenstein, I. et al. Factuality challenges in the era of large language models and opportunities for fact-checking. Nat. Mach. Intell. 6, 852–863 (2024).
Rosas-Chavoya, M., Gallardo-Salazar, J. L., López-Serrano, P. M., Alcántara-Concepción, P. C. & León-Miranda, A. K. QGIS a constantly growing free and open-source geospatial software contributing to scientific development. Cuad. Investig. Geogr. 48, 197–213 (2022).
Chrisman, N. Full circle: more than just social implications of GIS. Cartographica 40, 23–35 (2005).
Gao, L., Schulman, J. & Hilton, J. Scaling laws for reward model overoptimization. Int. Conf. Mach. Learn. pp 10835–10866 (2023).
Yao, S. et al. React: synergizing reasoning and acting in language models. Int. Conf. Learn. Represent. (2023).
Wei, J. et al. Chain-of-thought prompting elicits reasoning in large language models. Adv. Neural Inf. Process. Syst. 35, 24824–24837 (2022).
Schick, T. et al. Toolformer: language models can teach themselves to use tools. Adv. Neural Inf. Process. Syst. 36, 68539–68551 (2023).
Dong, Q. et al. A survey on in-context learning. In Proc. 2024 Conference on Empirical Methods in Natural Language Processing, pp 1107–1128 (2024).
Rocktäschel, P. et al. Retrieval-augmented generation for knowledge-intensive nlp tasks. Adv. Neural Inf. Process. Syst. 33, 9459–9474 (2020).
Chase, H. LangChain https://github.com/langchain-ai/langchain (2022).
Moura, J., Crew AI. https://github.com/crewAIInc/crewAI (2024).
Duckduckgo-search. https://github.com/deedy5/duckduckgo_search (2024),
Richardson, L. Beautiful Soup Documentation. https://readthedocs.org/projects/beautiful-soup-4/downloads/pdf/latest/ (2007).
Lundh, F., Ellis, M. Python Imaging Library (PIL). https://github.com/python-pillow/Pillow (2012).
Post 2022 Event Flood Behaviour Analysis—Tweed River, New South Wales Government, Department of Planning and Environment, Australia. https://flooddata.ses.nsw.gov.au/flood-projects/post-2022-event-flood-behaviour-analysis-tweed-river-report-only (2024).
Resilient and ready for the people of Queensland, Queensland Fire Department, Australia. https://www.fire.qld.gov.au/sites/default/files/2021-04/Response-Magazine-Dec2019_0.pdf (2019).
Acknowledgements
The authors acknowledge the Sydney Informatics Hub and the use of the University of Sydney’s high-performance computing cluster, Artemis. The authors also acknowledge support from the University of Sydney through the Digital Sciences Initiative programme and from the Australian Research Council (ARC) under the Discovery Project DP230100749. The work is supported in part by the National Natural Science Foundation of China (Grant No. 52479098).
Author information
Authors and Affiliations
Contributions
Z.C. contributed to the conceptualisation and data curation, developed the overall framework, conducted validation and visualisation and led the writing of the original draft and subsequent revisions. E.A.S. contributed to the conceptualisation and supported the formal analysis, investigation, methodological development and paper review and editing. S.J. contributed to the formal analysis, investigation and methodology, provided computational resources and funding acquisition and assisted with paper review and editing. L.S. supported the conceptualisation, formal analysis, investigation and methodology and provided funding, supervision and paper review and editing. D.D.C. contributed to the conceptualisation, formal analysis, investigation and methodology, oversaw project administration, provided funding and resources, offered supervision and assisted with paper review and editing.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Yu-Jun Zheng and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Chen, Z., Asadi Shamsabadi, E., Jiang, S. et al. Integration of large vision language models for efficient post-disaster damage assessment and reporting. Nat Commun (2026). https://doi.org/10.1038/s41467-025-68216-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-025-68216-z
