Elamur Research⌕
☰
home / papers / 2026 / GCA Framework: A GCC Countries–Grounded Dataset and Agentic Pipeline for Climate Decision Support

GCA Framework: A GCC Countries–Grounded Dataset and Agentic Pipeline for Climate Decision Support

Paper · 2026 · Muhammad Umer Sheikh, Khawar Shehzad, Salman Khan et al.

Summary

General-purpose LLMs handle climate questions about the Gulf Cooperation Council (GCC) states poorly. They lack regional knowledge, misread numbers in climate plots, and struggle to call geospatial or forecasting tools correctly. The authors address this with the GCA framework, which has two parts. The first is GCA-DS, a semi-automatically built multimodal QA dataset of roughly 200k items. Its text comes from Gulf government policies, NGO and international reports, academic papers and news on extreme events. It also includes city-level visual-temporal artifacts derived from reanalysis, air-quality, hydrology and climate-simulation products, with human verification applied to a subset. The second part is Gulf Climate Agent (GCA), a LoRA-fine-tuned Qwen2.5-VL 7B. It routes queries to a modular tool suite covering remote sensing, weather and rainfall, air quality and health, hydrology, carbon, retrieval and geocoding. On a tool-use benchmark, GCA far outperforms its untuned backbone and matches or slightly exceeds GPT-5, mostly because it produces fewer malformed tool calls. The work matters as the first Gulf-specific resource combining policy grounding, remote-sensing evidence and agentic tool use for climate decision support.

Key points

  • GCA-DS size and formats: GCA-DS contains 201,410 QA pairs, split into about 110k textual and about 90k visual-temporal items. Questions come in MCQ, open-ended and true/false formats and span all six GCC member states: the UAE, Saudi Arabia, Qatar, Oman, Kuwait and Bahrain.
  • Visual data sources: The visual split draws on ERA5/ERA5-Land, CAMS, GloFAS and CMIP6 HighResMIP. Sentinel-2 and Google Earth Engine supply remote-sensing inputs.
  • Benchmark: It includes 91 manually annotated questions. Step-wise metrics are InstAcc, ToolAcc, ArgAcc and SummAcc, and end-to-end metrics are AnsAcc and AnsAcc+I.
  • Main results: GCA reaches ToolAcc 94.2 versus 62.2 for base Qwen2.5-VL 7B. Its AnsAcc is 88.2, compared with 52.3 for the base model, 86.3 for GPT-5, 84.3 for Claude 4.5 Sonnet and 83.2 for Gemini 2.5 Pro.
  • Error reduction: Format errors fall to 12.7% from 46.7%, argument errors to 8.3% from 38.2%, and no-action cases to 3.9% from 21.5%.

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

🎓Mohamed bin Zayed University of Artificial IntelligenceAbu Dhabi · United Arab Emirates124 works · 5,676 citations🎓Australian National UniversityCanberra · Australia23 works · 6,522 citations🎓Linköping UniversityLinköping · Sweden18 works · 2,319 citations

Also on the byline: University of Missouri.

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksEnvironmental Science · 9 worksGlobal and Planetary Change · 1 work

Cite

Cite as Muhammad Umer Sheikh, Khawar Shehzad, Salman Khan, Fahad Shahbaz Khan and Muhammad Haris Khan. 2026. GCA Framework: A GCC Countries–Grounded Dataset and Agentic Pipeline for Climate Decision Support. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1967

@inproceedings{sheikh2026gca,
  title = {{GCA Framework: A GCC Countries–Grounded Dataset and Agentic Pipeline for Climate Decision Support}},
  author = {Muhammad Umer Sheikh, Khawar Shehzad, Salman Khan, Fahad Shahbaz Khan and Muhammad Haris Khan},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1967},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1967}
}

References

37 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sören Auer, Dante A. C. Barone, Cassiano Bartz, Eduardo G. Cortes, Mohamad Yaser Jaradeh, Oliver Karras, Dmitry Mouromtsev, and 1 others. 2023. The sciqa scientific question answering benchmark for scholarly knowledge. Scientific Reports, 13:7240. find
  2. Jannis Bulian, Mike S. Schäfer, Afra Amini, Heidi Lam, Massimiliano Ciaramita, Ben Gaiarin, Michelle Chen Huebscher, Christian Buck, Niels G. Mede, Markus Leippold, and Nadine Strauss. 2024. Assessing large language models on climate information. In Proceedings of the 41st International Conference o… find
  3. Jian Chen, Peilin Zhou, Yining Hua, Dading Chong, Meng Cao, Yaowei Li, Zixuan Yuan, Bing Zhu, and Junwei Liang. 2024. Vision-language models meet meteorology: Developing models for extreme weather events detection with heatmaps. Preprint, arXiv:2406.09838. — [arXiv:2406.09838](https://arxiv.org/abs/2406.09838) find
  4. Copernicus Atmosphere Monitoring Service (CAMS). 2026a. CAMS European air quality reanalyses. Copernicus Atmosphere Data Store (ADS). Accessed 2026-01-04. find
  5. Copernicus Atmosphere Monitoring Service (CAMS). 2026b. CAMS Global atmospheric composition forecasts. ECMWF datasets. Accessed 2026-01-04. find
  6. Copernicus Atmosphere Monitoring Service (CAMS). 2026c. CAMS global greenhouse gas forecasts. Copernicus Atmosphere Data Store (ADS). Accessed 2026-01-04. find
  7. Copernicus Atmosphere Monitoring Service (CAMS). 2026d. CAMS UV index forecasts. Copernicus Atmosphere Monitoring Service (CAMS) charts. Accessed 2026-01-04. find
  8. Copernicus Atmosphere Monitoring Service (CAMS). 2026e. Ground-level pollen forecast (CAMS). European Climate and Health Observatory / Climate-ADAPT. Accessed 2026-01-04. find
  9. Copernicus Climate Change Service (C3S). 2026a. ERA5 hourly data on single levels from 1940 to present. Copernicus Climate Data Store (C3S). Accessed 2026-01-04. find
  10. Copernicus Emergency Management Service (CEMS). 2026b. GloFAS v4.0 basics: hydrological model and implementation set-up. Copernicus Emergency Management Service (CEMS) / ECMWF Confluence. Accessed 2026-01-04. find
  11. Copernicus Emergency Management Service (CEMS). 2026c. GloFAS v4.0 hydrological reanalysis. Joint Research Centre (JRC) Data Catalogue / Copernicus Emergency Management Service. Accessed 2026-01-04. find
  12. European Centre for Medium-Range Weather Forecasts (ECMWF). 2026. ECMWF Reanalysis v5 (ERA5). ECMWF dataset description. Accessed 2026-01-04. find
  13. Zhicheng Guo, Sijie Cheng, Hao Wang, Shihao Liang, Yujia Qin, Peng Li, Zhiyuan Liu, Maosong Sun, and Yang Liu. 2024. Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 11143… find
  14. R. J. Haarsma, M. J. Roberts, P. L. Vidale, C. A. Senior, A. Bellucci, Q. Bao, P. Chang, and 1 others. 2016. High resolution model intercomparison project (highresmip v1.0) for cmip6. Geoscientific Model Development, 9:4185–4208. find
  15. Bruce Hewitson, Anthony C. Janetos, Timothy R. Carter, Filippo Giorgi, Richard G. Jones, Won-Tae Kwon, Linda O. Mearns, E. Lisa F. Schipper, and Maarten K. van Aalst. 2014. Regional context. In V. R. Barros, C. B. Field, D. J. Dokken, M. D. Mastrandrea, K. J. Mach, T. E. Bilir, M. Chatterjee, K. L.… find
  16. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen. 2022. Lora: Low-rank adaptation of large language models. In International Conference on Learning Representations (ICLR).
  17. Copernicus Climate Change Service (C3S). 2026b. ERA5-Land hourly data from 1950 to present. Copernicus Climate Data Store (C3S). Accessed 2026-01-04. find
  18. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP).
  19. Copernicus Emergency Management Service (CEMS). 2026a. GloFAS forecasts: product technical information. Copernicus Emergency Management Service (CEMS) Global Flood Awareness System. Accessed 2026-01-04. find
  20. Hyeonjae Kim, Chenyue Li, Wen Deng, Mengxi Jin, Wen Huang, Mengqian Lu, and Binhang Yuan. 2025. Climateagent: Multi-agent orchestration for complex climate data science workflows. Preprint, arXiv:2511.20109. — [arXiv:2511.20109](https://arxiv.org/abs/2511.20109) find
  21. Murathan Kurfali, Shorouq Zahra, Joakim Nivre, and Gabriele Messori. 2025. Climateeval: A comprehensive benchmark for nlp tasks related to climate change. In Proceedings of the 2nd Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2025), Vienna, Austria. Association for Comput… find
  22. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval-augmented generation for knowledge-intensive NLP tasks. In Advances in Neural Information Proc…
  23. Xiao Liu, Hao Yu, Hanchen Zhang, Yifan Xu, Xuanyu Lei, Hanyu Lai, Yu Gu, Hangliang Ding, Kaiwen Men, Kejuan Yang, Shudan Zhang, Xiang Deng, Aohan Zeng, Zhengxiao Du, Chenhui Zhang, Sheng Shen, Tianjun Zhang, Yu Su, Huan Sun, and 3 others. 2024. Agentbench: Evaluating llms as agents. In International…
  24. Pan Lu, Swaroop Mishra, Tony Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Øyvind Tafjord, Peter Clark, and Ashwin Kalyan. 2022. Learn to explain: Multimodal reasoning via thought chains for science question answering. In Advances in Neural Information Processing Systems (NeurIPS). find
  25. Veeramakali Vignesh Manivannan, Yasaman Jafari, Srikar Eranky, Spencer Ho, Rose Yu, Duncan Watson-Parris, Yian Ma, Leon Bergen, and Taylor Berg-Kirkpatrick. 2025. Climaqa: An automated evaluation framework for climate question answering models. In International Conference on Learning Representations… find
  26. Ahmed Masry, Do Xuan Long, Jia Qing Tan, Shafiq Joty, and Enamul Hoque. 2022. Chartqa: A benchmark for question answering about charts with visual and logical reasoning. In Findings of the Association for Computational Linguistics: ACL 2022, pages 2263– 2279, Dublin, Ireland. Association for Computa…
  27. Michael D. Mastrandrea, Katharine J. Mach, Gian-Kasper Plattner, Ottmar Edenhofer, Thomas F. Stocker, Christopher B. Field, Kristie L. Ebi, and Patrick R. Matschoss. 2011. The IPCC AR5 guidance note on consistent treatment of uncertainties: a common approach across the working groups. Climatic Chang… find
  28. Nitesh Methani, Pritha Ganguly, Mitesh M. Khapra, and Pratyush Kumar. 2020. Plotqa: Reasoning over scientific plots. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 1527–1536. find
  29. Climate paper image question answering dataset for retrieval-augmented generation with context-based query expansion. In Proceedings of the 2nd Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2025), pages 218–232, Vienna, Austria. Association for Computational Linguistics. find
  30. Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, Sihan Zhao, Lauren Hong, Runchu Tian, Ruobing Xie, Jie Zhou, Mark Gerstein, Dahai Li, Zhiyuan Liu, and Maosong Sun. 2023. Toolllm: Facilitating large language models to master 16000+ real…
  31. Md Rashad Al Hasan Rony, Ying Zuo, Liubov Kovriguina, Roman Teucher, and Jens Lehmann. 2022. Climate bot: A machine reading comprehension system for climate change question answering. In Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence (IJCAI-22), AI for Good… find
  32. Muhammad Umer Sheikh, Hassan Abid, Khawar Shehzad, Ufaq Khan, and Muhammad Haris Khan. 2025. Mmclima: A framework for multimodal climate science data and evaluation. https://openreview.net/forum?id=j9TdFswuZ3. OpenReview preprint. ↗
  33. Yuwei Wan, Yixuan Liu, Aswathy Ajith, Clara Grazian, Bram Hoex, Wenjie Zhang, Chunyu Kit, Tong Xie, and Ian Foster. 2024. Sciqag: A framework for auto-generated science question answering dataset with fine-grained evaluation. Preprint, arXiv:2405.09939. — [arXiv:2405.09939](https://arxiv.org/abs/2405.09939) find
  34. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2023. Self-instruct: Aligning language models with self-generated instructions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL). find
  35. Qinzhuo Wu, Wei Liu, Jian Luan, and Bin Wang. 2024. Toolplanner: A tool augmented llm for multi granularity instructions with path planning and feedback. Preprint, arXiv:2409.14826. find
  36. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR).
  37. Hongyin Zhu and Prayag Tiwari. 2023. Climate change from large language models. Preprint, arXiv:2312.11985. — [arXiv:2312.11985](https://arxiv.org/abs/2312.11985) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/sheikh-2026-gca-framework-gcc-countries-grounded/