Climate

Experts: How are climate scientists using artificial intelligence?

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 22/10/2026, 11:05:24 UTC.

Briefings

Climate scientists are utilizing artificial intelligence to enhance weather forecasting and climate research, leveraging large language models and machine learning to process vast datasets. Experts note that while AI excels at short-term predictions using available data, it struggles with extreme events and climate projections beyond observed records. Key findings include the UK Met Office's use of AI for near-term forecasts requiring less computing power, and the University of East Anglia's recognition of AI's role in numerical weather prediction. However, significant gaps remain regarding AI's performance on unprecedented extremes and the challenges of predicting future climates without sufficient historical data.

Key points

• AI weather-prediction models at the UK Met Office utilize large language models to generate forecasts up to two weeks ahead using far less computing power than traditional numerical models.
• AI systems at the European Center for Mid-Range Weather Forecasts are now deployed alongside traditional systems for official weather communication, demonstrating competitive accuracy with leading numerical models.

Context

Climate scientists are employing artificial intelligence to accelerate research and improve forecasting accuracy. This includes using large language models for data analysis and machine learning for numerical weather prediction, allowing for more efficient study of the warming planet.

Limitations

Evidence is limited to interviews with 10 experts covering specific themes. Uncertainty exists regarding AI performance on unprecedented extremes and climate projections beyond observed records. The field lacks sufficient historical data to fully train models for future climates, and the central open problem remains solving this data gap.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 22/10/2026, 11:11:22 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeicdtjtlzhev2qezz5kctsqpbvv2stv3kqkdz4zoz5tr6wd6hrgzse