通过转移学习改进观测稀疏区域的海洋再分析
Simon Lentz1, Sebastian Brune2, Christopher Kadow3
1Institute of Oceanography, Center for Earth System Sustainability, Universität Hamburg, Hamburg, Germany. simon.lentz@uni-hamburg.de.
转移学习神经网络改善了稀疏的海洋地下温度重建. 这种机器学习方法增强了气候障碍预测,并显示了气候科学之外的混合频率数据分析的潜力.
科学领域:
- 海洋学 海洋学 海洋学
- 气候科学 气候科学
- 机器学习 机器学习
背景情况:
- 海洋地下的观测是有限的,导致基于模型的估计存在重大不确定性.
- 准确的历史海洋温度数据对于了解气候变化和改善气候模型至关重要.
研究的目的:
- 调查基于转移学习的神经网络的有效性,以重建北大西洋地下温度,在稀少观测数据的时期.
- 评估这些网络将学习的物理模式转移到历史海洋热量含量估计中的能力.
主要方法:
- 在海洋观测丰富的时期训练神经网络以学习物理行为.
- 在数据同化框架内评估网络,以确保其重现已学习的物理模式.
- 应用训练网络来重建使用不常见输入数据的温度.
主要成果:
- 转移学习网络成功地从训练数据中学习和复制物理模式.
- 使用不常见数据的重建显示了相似的物理结构,并与最先进的方法相比纠正了已知的错误.
- 该网络证明了结果从高频到低频的准确传输.
结论:
- 转移学习神经网络提供了一种可行的方法,以稀疏的数据增强海洋地下温度重建.
- 这种方法可以改善气候后预测的初始化和评估.
- 该技术在气候科学和其他领域涉及混合频度测量的应用方面表现有前途.
更多相关视频
10:28Investigating the Relationship between Sea Surface Chlorophyll and Major Features of the South China Sea with Satellite Information
Published on: June 13, 2020
08:15Author Spotlight: Unveiling Plankton Response to Climate Change Through Time-Series Data and Artistic Expression
Published on: July 28, 2023
相关概念视频
Precipitation Gravimetry
In determining nickel by gravimetric analysis, a precipitant of ethanolic dimethylglyoxime is added to a hot nickel salt solution. This is quickly followed by the dropwise addition of dilute ammonia solution until precipitation occurs. A...
Improving Translational Accuracy
Region of Convergence of Laplace Tarnsform
Consider a decaying exponential signal that begins at a specific time. When deriving its Laplace transform, the time-domain variable is replaced with a complex variable. This...
Methods of Obtaining Topography
Residuals and Least-Squares Property
If the observed data point lies above the line, the residual is positive, and the line underestimates the actual data value for y. If the observed data point lies below the line, the residual is negative, and the line overestimates the actual data value for y.
The process of fitting the best-fit...
Stratified Sampling Method
To choose a stratified sample, divide the population into groups called strata and then take a...
