Lunch seminar: LLM-driven approach to critical climate adaptation challenges
WaterCentre@KTH proudly presents a lunch seminar with Haozhi Pan, Professor at the School of International and Public Affairs, Shanghai Jiao Tong University.
Time: Wed 2026-02-25 12.00 - 13.00
Location: Climate Action House, Teknikringen 43
Language: English
Participating: Haozhi Pan, Professor at the School of International and Public Affairs, Shanghai Jiao Tong Universi
This seminar presents results from two connected studies using LLM-driven methods to address key climate adaptation challenges.
The first study applies multilingual LLMs to over 2.1 million news articles from 184 countries to identify global reporting gaps on extreme heat. Combined with climate modeling and human-in-the-loop validation, the analysis shows that highly vulnerable but under-reported regions – such as equatorial Africa, South Asia, and polar and high-mountain areas – face substantially higher economic losses and mortality. Under SSP5-8.5, projected heat-related losses in these regions reach 36% above the global average by 2050, highlighting how data and language gaps exacerbate vulnerability and misallocate adaptation resources.
The second study uses an LLM + Retrieval-Augmented Generation (RAG) framework to analyse more than 2,000 historical transboundary water cooperation cases (1951–2019). By integrating treaty texts, reports, news, and hydrological and socioeconomic data, it identifies six cooperation archetypes and shows that integrated basin agreements and joint water allocation mechanisms are most effective at reducing conflict. Projections suggest climate change could double conflict risk in several global hotspots by 2050, while proactive, strategy-informed cooperation could reduce over half of this risk, offering clear guidance for strengthening resilience in shared river basins.
The first 15 to arrive will get a vegetarian wrap and a soft drink!