Collaborative Research: GEM--Towards Developing Physics-informed Subgrid Models for Geospace MagnetoHydroDynamics (MHD) Simulations

在模拟太阳风和磁层系统之间的相互作用时，科学家通常使用数值磁流体动力学（MHD），这是一种导电流体模型，将所有相互穿透的粒子物种一起视为单个连续介质。越来越多的MHD模型需要非常高的数值分辨率来模拟多尺度等离子体流的真实全球磁层。为了解决这个问题，该项目将利用物理信息机器学习和随机建模的数据驱动发现，为现有的全球磁层MHD模型开发新的参数化。该项目将支持一位早期职业科学家担任高级职位。主要的更广泛的影响将是改进全球磁层模型的MHD分量，从而更好地模拟和预测空间天气。所开发的技术是非常通用的，可以适用于其他复杂的高维动力系统，有利于其他科学和工程领域。为了扩大结果并证明其鲁棒性，GAMERA将采用一系列物理问题对几种类型的多尺度湍流MHD流动进行动力学模拟，并通过增加参考数据复杂性来系统地提升：(1)Orszag-Tang涡旋的二维模拟，(2)Kelvin-Helmholtz不稳定性的二维模拟，(3)近地磁尾中突发体流的三维模拟。从基准的高分辨率GAMERA模型解中诊断出以下关键的时空参考数据：(i)子网格（小尺度）和大尺度场的分布，（ii）封装大尺度场诱导反馈的子网格尺度强迫。基于物理的机器学习和随机建模将用于开发亚电网尺度和诱导强迫的预测模型，这些模型与粗尺度GAMERA模拟的大尺度流相结合。已开发的子网格尺度参数化的技能将通过与实际应用相关的综合物理信息指标集进行正式和系统的评估。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

While simulating the interaction between the solar wind and magnetosphere system, scientists usually use numerical magnetohydrodynamics (MHD), a model of electrically conducting fluids that treats all interpenetrating particle species together as a single continuous medium. Increasingly, MHD models require very-high numerical resolution for realistic global magnetosphere simulations of multiscale plasma flows. To address this problem, this project will develop new parameterizations for an existing global magnetosphere MHD model with the data-driven discovery by physics-informed machine learning and stochastic modeling. The project will support an earlier career scientist in a senior personnel role. The main broader impact will be the improvement of MHD components of global magnetosphere models, leading to better modeling and prediction of space weather. The developed techniques are very general and can be adapted to other complex high-dimensional dynamical systems with benefits to other areas of science and engineering. To broaden the results and prove their robustness, a hierarchy of physical problems will be employed for dynamical simulations of several types of multiscale turbulent MHD flows by GAMERA, to ascend systematically by increasing the reference data complexity: (1) 2D simulation of Orszag-Tang vortex, (2) 2D simulation of the Kelvin-Helmholtz instability, (3) 3D simulation of bursty bulk flows in the near-Earth magnetotail. The following key spatiotemporal reference data will be diagnosed from benchmark high-resolution GAMERA model solutions: (i) distributions of subgrid (small-scale) and large-scale fields, (ii) subgrid-scale forcing that encapsulates induced feedbacks on the large-scale fields. Physics-informed machine learning and stochastic modeling will be used to develop prognostic models of subgrid-scales and induced forcing, coupled to large-scale flow simulated by the coarse-scale GAMERA. Skills of the developed subgrid-scale parameterizations will be formally and systematically evaluated by the comprehensive set of physics-informed metrics relevant to practical applications.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.