Variational inference of ice shelf rheology with physics-informed machine learning
Variational inference of ice shelf rheology with physics-informed machine learning
复制标题
DOI：
10.1017/jog.2023.8
复制
发表时间：
2023-04
影响因子：
3.4
通讯作者：
B. Riel;B. Minchew
中科院分区：
文献类型：
--
作者：
B. Riel;B. Minchew
摘要 南极洲海岸边缘的漂浮冰架阻止了接地冰流入海洋。控制冰架提供的流动阻力的关键因素之一是构成冰架的冰的刚性（与粘度相关）。冰的硬度是高度异质的，必须通过空间连续的表面观测并同化到冰流模型中进行校准。需要校准刚性值的现实不确定性来量化冰盖和海平面预测的不确定性。在这里，我们提出了一个基于物理的机器学习框架，用于推断给定冰架的刚性值的完整概率分布，以从遥感数据导出的冰表面速度和厚度场为条件。我们采用变分推理来联合训练神经网络和变分高斯过程来重建表面观测值、刚性值和不确定性。将框架应用于南极洲的合成冰架和大型冰架表明，在观测的噪声水平内可以测量冰变形时，刚性受到很好的约束。通过用传统反演方法补充变分推理可以进一步降低不确定性。我们的结果展示了通过遥感观测不断更新冰流参数校准的前进道路。
Abstract Floating ice shelves that fringe the coast of Antarctica resist the flow of grounded ice into the ocean. One of the key factors governing the amount of flow resistance an ice shelf provides is the rigidity (related to viscosity) of the ice that constitutes it. Ice rigidity is highly heterogeneous and must be calibrated from spatially continuous surface observations assimilated into an ice-flow model. Realistic uncertainties in calibrated rigidity values are needed to quantify uncertainties in ice sheet and sea-level forecasts. Here, we present a physics-informed machine learning framework for inferring the full probability distribution of rigidity values for a given ice shelf, conditioned on ice surface velocity and thickness fields derived from remote-sensing data. We employ variational inference to jointly train neural networks and a variational Gaussian Process to reconstruct surface observations, rigidity values and uncertainties. Applying the framework to synthetic and large ice shelves in Antarctica demonstrates that rigidity is well-constrained where ice deformation is measurable within the noise level of the observations. Further reduction in uncertainties can be achieved by complementing variational inference with conventional inversion methods. Our results demonstrate a path forward for continuously updated calibrations of ice flow parameters from remote-sensing observations.