研究人员利用生成式人工智能模拟了本世纪气候变化的数百种可能情景，以研究极端降雨在全球变暖背景下如何加剧1。通过训练140年区域气候模拟数据，AI在6小时内生成了超过15000年的模拟数据，相比传统模型需耗时数年，计算效率实现了显著提升1。
研究成果表明，新西兰百年一遇的极端降水事件随全球温度每升高1°C而增加约9%，远超常见降水事件5.6%的增速1。该研究基于20个全球气候模型的数据，采用约12公里宽的网格精度来检查新西兰的区域差异1。
Researchers have employed generative artificial intelligence to simulate hundreds of potential climate change scenarios this century, examining how extreme rainfall will intensify as global temperatures rise.1 The study found that century-scale extreme precipitation events in New Zealand increase by approximately 9% for every 1°C of global warming, significantly outpacing the 5.6% increase observed in common rainfall events.1
The AI model was trained on 140 years of regional climate simulation data spanning both historical and warmer future conditions, enabling it to generate over 15,000 years of simulated data within six hours using four specialized computing chips.1 This computational efficiency represents a dramatic improvement over traditional climate models, which require years to complete equivalent calculations.1 The simulations operated at a grid resolution of approximately 12 kilometers to capture regional variations across New Zealand, drawing data from 20 global climate models.1
评论
还没有评论，欢迎留下第一条。