通过方程发现和自动调整,在混合AI气候模型中减少云覆盖错误
Arthur Grundner1, Tom Beucler2,3, Julien Savre4
1Institut für Physik der Atmosphäre, Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR), Oberpfaffenhofen, Germany. arthur.grundner@dlr.de.
这项研究为气候模型引入了可解释的机器学习,改善了云层覆盖预测. 改进的模型减少了偏差,在变暖的情况下保持准确,加强了地球系统模型 (ESM) 的忠实性.
科学领域:
- 气候科学 气候科学
- 大气物理学 大气物理学
- 机器学习应用 机器学习应用
背景情况:
- 云的参数化是气候预测中不确定性的主要来源.
- 对于地球系统模型 (ESM) 的现有数据驱动机器学习方法往往缺乏可解释性和物理一致性.
研究的目的:
- 在全球大气模型中开发和实施云层覆盖的物理一致,可解释的机器学习参数化.
- 通过加强ESM来提高气候预测的准确性和减少偏差.
主要方法:
- 在ICON全球大气模型中使用符号回归来进行风暴解决模拟的物理一致的云盖参数化.
- 应用了Nelder-Mead优化器,以在嵌套阶段对地球观测进行混合模型自动重新校准.
主要成果:
- 混合模型显著减少了云层覆盖的偏差,在南大洋上减少了75%,在亚热带层积地区减少了44%.
- 改进后的模型在+4K表面变暖条件下表现出强性.
结论:
- 可解释的机器学习参数化,结合实际的调整方法,可以有效地提高地球系统模型的可靠性.
- 这种方法提供了一种透明和有效的方式来加强气候预测.
更多相关视频
12:11Computation of Atmospheric Concentrations of Molecular Clusters from ab initio Thermochemistry
Published on: April 8, 2020
10:36Author Spotlight: Optimization of Airflow Velocities in Battery Cooling Systems for Enhanced Thermal Performance and Reduced Energy Consumption
Published on: November 3, 2023
相关概念视频
Precipitation Processes
Maxwell-Boltzmann Distribution: Problem Solving
This distribution function f(v) is defined by saying that the expected number N (v1,v2) of particles with speeds between v1 and v2 is given by
Variation of Atmospheric Pressure
Assuming the air temperature is constant at a given altitude and that the ideal gas law of thermodynamics describes the atmosphere to a good approximation, one can find the variation of atmospheric pressure with height.
Let p(y) be the atmospheric pressure at...
Distribution Reliability and Automation
Heating and Cooling Curves
For instance, the addition of heat raises the temperature of a solid; the amount of heat absorbed depends on the heat capacity of the solid (q = mcsolidΔT). According to thermochemistry, the relation between the amount of heat absorbed or released by a substance, q, and its...
Precipitation and Co-precipitation
