Collaborative Research: The Land Unknown: Assessing Data Requirements for Modeling Change in the Antarctic Ice Sheet with an Emphasis on the Subglacial Bed

极地冰冻圈目前正在发生的最戏剧性的变化之一是格陵兰岛和西南极出口冰川的几倍加速。冰盖的计算模型在正确模拟这些事件方面面临着相当大的挑战，这反过来限制了冰川学家和气候学家预测未来变化的能力。这一挑战源于对关键变量的有限了解，例如冰下陆地表面的形状。数据限制带来了两个问题：第一，可能会遗漏冰行为的重要细节；第二，模型可能会产生与观测结果相当好的结果，但这样做的原因是错误的。该项目的目标是将这些不确定性本身作为指南，以确定哪些类型的数据对产生对未来冰盖变化的有用预测最重要。这项工作主要集中在两个地区：东南极的奥罗拉盆地，最近在那里可以获得新的高分辨率数据，以及位于南极腹地或东南极深处的南极站周围的集水区，那里的冰流目前缓慢，过去可能会更快。贝叶斯推理可用于通过合成来自不同数据源的信息、建模结果和相应的误差信息来正式调查各种模型属性对模型投影中的不确定性的贡献。在这里，冰盖模型将被要求根据适当的边界条件和初始模型状态来确定对冰流和厚度变化影响最大的区域和过程。这种方法将使我们能够创建和分析相关图，显示选定研究区域中各种边界条件和模型结果之间的关系。那些最重要的参数或区域应该主导变化模型预测中的分散。相关的后验概率密度将为如何设计观测策略以满足特定的科学标准提供指导。

Among the most dramatic changes now underway in the polar cryosphere is the several-fold speedup of Greenland and West Antarctic outlet glaciers. Computational models of ice sheets face considerable challenges in correctly simulating these events, which in turn, limits glaciologists' and climatologists' ability to predict future change. The challenge arises from limited knowledge of key variables, for example, the shape of the land surface beneath the ice. Data limitations present two problems; first, important details of ice behavior may be missed and second, models may produce results that compare well with observations but do so for the wrong reasons. The goal of this project is to use these uncertainties themselves as a guide to identifying which types of data are most important to producing useful projection of future ice sheet change. The work is focused on two regions, the Aurora basin in East Antarctica, where new high resolution data have recently become available, and the catchment surrounding South Pole station deep in the interior or East Antarctica, where ice flow is currently sluggish may have been faster in the past. Bayesian inference may be used to investigate formally the contributions of various model attributes to uncertainty in model projections by synthesizing information from different sources of data, modeling results, and corresponding error information. Here, an ice sheet model will be asked to identify the regions and processes that have the largest influence on ice flow and thickness change, according to an appropriate range of boundary conditions and initial model states. The approach will allow us to create and analyze correlation maps showing relationships between various boundary conditions and model outcome in selected study areas. Those parameters or regions that are most important should dominate the scatter in model projections of change. The associated posterior probability density will provide guidance on how to design observational strategies to meet specified scientific criteria.