研究人员在备受欢迎的变压器模型(名为三维地质模型)的基础上，开发了一种用于厄尔尼诺-南方涛动预测的特定自注意神经网络模型，该模型用于预测三维(3D)海洋上层温度异常和风应力异常。这种纯数据驱动和时空注意力增强的模型对于提前18个月进行并在北方春季开始启动的Nino 3.4海温异常预测中，实现了惊人的高关联技能。此外，灵敏度实验表明，三维地质模型可以描述厄尔尼诺-南方涛动周期中Bjerknes反馈机制下的海洋上层温度演化和海洋-大气耦合动力学。基于自我注意力模型在厄尔尼诺-南方涛动预测中的成功实现，表明其在地学多维时空建模方面具有巨大潜力。
据悉，使用基于过程的动力学模型对厄尔尼诺-南方涛动(ENSO)的实时预测仍然存在很大的偏差和不确定性。数据驱动的深度学习算法的最新进展为热带太平洋海面温度(SST)建模提供了一种有前途的方法。
附：英文原文
Title: A self-attention–based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions
Author: Lu Zhou, Rong-Hua Zhang
Issue&Volume: 2023-03-08
Abstract: Large biases and uncertainties remain in real-time predictions of El Nio–Southern Oscillation (ENSO) using process-based dynamical models; recent advances in data-driven deep learning algorithms provide a promising mean to achieve superior skill in the tropical Pacific sea surface temperature (SST) modeling. Here, a specific self-attention–based neural network model is developed for ENSO predictions based on the much sought-after Transformer model, named 3D-Geoformer, which is used to predict three-dimensional (3D) upper-ocean temperature anomalies and wind stress anomalies. This purely data-driven and time-space attention-enhanced model achieves surprisingly high correlation skills for Nino 3.4 SST anomaly predictions made 18 months in advance and initiated beginning in boreal spring. Further, sensitivity experiments demonstrate that the 3D-Geoformer model can depict the evolution of upper-ocean temperature and the coupled ocean-atmosphere dynamics following the Bjerknes feedback mechanism during ENSO cycles. Such successful realizations of the self-attention–based model in ENSO predictions indicate its great potential for multidimensional spatiotemporal modeling in geoscience.
DOI: adf2827
Source: https://www.science.org/doi/10.1126/sciadv.adf2827