机器学习模拟千年气候

Fountain: AGU Advances This is an authorized translation of an Eos article. 本文是Eos文章的授权翻译. 近年来,科学家们发现,基于机器学习的天气模型可以比传统模型更快地做出天气预测,且并且到第60天时就会开始模拟出不切实际的天气. 深度学习地球系统模型(Deep Learning Earth System Model,简称DLESyM)建立在两个并行运行的神经网络上:一个模拟海洋,另一个模拟大气。在模式运行期间,对海洋状况的预测每四个模式日更新一次。由于大气条件演变得更快,对大气的预测每12个模式小时更新一次。 该模型的创建者Cresswell-Clay 等人发现,DLESyM与过去观测到的气候非常吻合,并能做出准确的短期预测。以地球当前的气候为基准,它还可以在不到12小时的计算时间内,准确模拟1000式比对计划第六阶段(CMIP6)的模型相当,甚至优于后者,CMIP6目前在计算气候研究中被广泛使用. DLESyM 模型在模拟热带气旋和印度夏季季风方面优于 CMIP6 模型。它至少与CMIP6 1000年模拟结束时(3016 年)生成的东北风暴的结构与 2018 年观测到的东北风暴非常相似. 然而,新模型和CMIP6此外,对于中期预报(即未来 15天左右的预报),DLESyM DLESyM DLESyM compliant, CMIP6 compliant (AGU Advances, https://doi.org/10.1029/2025AV0017062025) —科学撰稿人Madeline Reinsel This translation was done by wiley. 本文翻译由wiley提供。 Read this article on WeChat. 在微信上阅读本文. Text…

Read More

The Big Battery Snowy 2 pumping hydroelectric project in Australia suffers from another cost explosion: Watts agrees with that?

Eric World essay One billion here, one billion there, you will soon be talking about real money. Snowy Hydro 2.0 indicates another cost explosion with a price of 12 billion dollars that is now considered unattainable The pumping hydroelectric project in the Kosciuszko National Park is complete at 67%, but the cost increases in the…

Read More

Faced with the decimation of federal science, our community defends science as Never Before – The Bridge: Connecting Science and Policy

“I came today to put a face to science and what these cuts mean, and I also wanted to change the script to be called ‘scammer’ to be called a patriot. I think scientists serve their country using the skills they have.” –Dr Jennifer Moraash, June 2025 days of participating action Scientists serve communities through…

Read More