ClimART
收藏资源简介:
ClimART数据集是由达姆施塔特工业大学和米拉共同创建的,包含超过1000万个样本,涵盖当前、工业化前和未来气候条件。该数据集基于加拿大地球系统模型,旨在通过机器学习方法模拟和加速天气及气候模型中的大气辐射传输计算。ClimART数据集不仅提供了丰富的气候数据,还为机器学习社区提供了多样的测试集和物理信息,以促进模型在不同气候条件下的泛化能力。此外,数据集还关注了准确性与推理速度之间的平衡,为气候科学中的机器学习应用提供了重要资源。
ClimART dataset was co-developed by Technische Universität Darmstadt and Mila, containing over 10 million samples covering current, pre-industrial, and future climate conditions. Based on the Canadian Earth System Model, this dataset aims to simulate and accelerate atmospheric radiative transfer calculations in weather and climate models via machine learning methods. The ClimART dataset not only provides abundant climate data, but also offers diverse test sets and physical information for the machine learning community, to enhance the generalization capability of models across different climate conditions. Furthermore, the dataset prioritizes the balance between accuracy and inference speed, serving as a critical resource for machine learning applications in climate science.

- 1ClimART: A Benchmark Dataset for Emulating Atmospheric Radiative Transfer in Weather and Climate Models达姆施塔特工业大学 & 米拉 · 2021年



