研究人员正在开发利用机器学习检测火山周围微弱振动变化的技术，以实现火山喷发前数小时的提前预警。1在新西兰怀卡里火山的测试中，这套预报器成功预测了五次喷发中的四次，但代价是每年约15天的虚假警报。1尽管早期预警会增加误报天数，研究表明通过成本-损失模型分析，及时的预警和疏散行动能减少可预防损失的30-90%，因此接受更多虚报成为一种值得的安全权衡。1
火山突发喷发的威胁已有明确记录。12019年12月的怀卡里火山喷发造成22人死亡、25人重伤，而日本2014年御岳山喷发则导致63人丧生。1过去20年，新西兰经历了约半打的突发火山爆炸。1联合国已将目标定为到2027年为所有地区提供通用多灾害预警系统。1
Researchers have developed a machine learning approach to detect subtle vibrations around volcanoes, potentially enabling warnings several hours before an eruption occurs.1 Testing this technology at Whakaari, also known as White Island in New Zealand, the predictive system successfully forecast four out of five eruptions, though this early detection capability came at the cost of approximately 15 false alarm days per year.1
The trade-off between early warnings and false alarms appears justified by the potential to save lives. Cost-loss analysis demonstrates that preventive action and evacuation could reduce preventable losses by 30 to 90 percent.1 This safety benefit becomes particularly significant given historical volcanic disasters in the region: the 2019 Whakaari eruption killed 22 people and seriously injured 25 others,1 while New Zealand has experienced roughly half a dozen sudden volcanic explosions over the past two decades.1 Internationally, the 2014 eruption of Mount Ontake in Japan claimed 63 lives.1
The development of such early warning systems aligns with broader disaster preparedness goals, as the United Nations aims to provide universal multi-hazard early warning systems to all regions by 2027.1
评论
还没有评论，欢迎留下第一条。