Scientists Build AI Tools to Make Climate Data Easier to Use

Scientists in the United Kingdom have created two new artificial intelligence tools to help people understand climate data. These tools use a type of AI called large language models, which can read and respond to questions in plain English. The goal is to make important climate information available to more people — not just trained scientists. The tools were built with help from the UK Met Office, Britain's national weather and climate service.

For years, the UK Met Office has kept a large collection of climate data called UKCP18. This dataset shows how the UK's climate is likely to change all the way until the year 2100. It includes scientific reports, charts, graphs, and huge amounts of numbers from computer climate models. The problem is that most people found it very hard to use without help from a climate expert.

To fix this, researchers built an AI chatbot that lets users ask questions in everyday language. Instead of digging through technical reports, a person can simply type a question and get a clear answer. The chatbot can even pull up charts or graphs to go along with its response. If a question is unclear, the system asks the user to provide more details before giving an answer.

The chatbot uses a method called retrieval-augmented generation, or RAG. This means the AI looks up real information from the UKCP18 database before it answers. This helps prevent the AI from making up facts — a problem sometimes called 'hallucination.' By grounding its answers in real data, the chatbot gives users responses they can trust.

The second AI tool works differently. It acts like a recipe generator for climate science. When a climate expert needs to figure out how to use climate data for a specific situation — like planning for floods or heat waves — this tool builds a step-by-step plan, or 'recipe,' for them. The tool pulls information from a large map of connected ideas called a knowledge graph, which links climate variables, hazards, and key datasets together.

This second tool uses several AI 'agents' that each handle a different part of the job. One agent plans the workflow, another writes a draft recipe, and another searches for extra information. A lead agent oversees all of them and puts the final answer together. This teamwork between AI agents helps climate consultants do their jobs faster and at lower cost.

Both tools were designed with real users in mind. Scientists and experts at the UK Met Office helped shape and test the systems. The tools also include a 'human-in-the-loop' feature, which means a real person can step in to check or correct the AI's work at key moments. This helps make sure the tools stay accurate and trustworthy.

Researchers believe these tools could change how climate services are delivered around the world. By making data easier to access, more communities could use climate information to plan for the future. Still, experts warn that AI tools must be carefully checked before they are widely trusted. The work is seen as an important step toward making climate science useful for everyone.