AI-Assisted Climate Risk Research

AI-assisted climate risk research

ClimateNode is building tools for AI-assisted climate risk research. ClimateNode's goal is to routinely scan scientific papers, world news and reports^* for granular information about climate-related and other environmental hazards, and how they affect specific places, companies, assets and sectors. It is using Natural Language Processing tools (including large language models) and knowledge graphs to achieve this.

This information can be used to complement quantitative methods and build up a more complete picture of how climate risk is evolving in a location, including how populations are adapting, how hazards may interact, and how non-meteorological factors (such as urbanisation) are contributing to overall risk.

These techniques can be used to:

• compress knowledge dispersed amongst thousands of different documents
• horizon scan for risks relevant to individual companies and government agencies and their assets
• identify risks and historic impacts relevant to critical points in supply chains
• improve knowledge of secondary perils
• improve understanding of how drought, heat, floods and hail are affecting agriculture
• understand on-the-ground non-climate factors relevant to identifying risks, for example, water management practices

Illustrative Example: China

The map on the left shows an illustrative sample of information collated so far for selected places in China.

The map popups contain a mixture of information which has been derived using OpenAI’s GPT-3.5-Turbo model to interrogate relevant documents^* (with some editing), and information which has been entered manually (for rights reasons).