A novel approach to constraining ice sheet models with glacial geomorphology

格陵兰岛和南极洲的冰盖容纳的水足以使全球海平面上升约65米，目前正在以加速的速度失去质量。对未来质量损失的预测是困难的，这意味着冰盖的贡献是估计未来海平面上升的一大不确定性来源。做出这些预测的数值模型需要对冰盖行为的观察来验证和改进，但现有冰盖的记录很短。通过研究上一次冰期(约12万至约10000年前)冰盖占据的区域，可以获得更长的冰盖行为记录。在此期间存在的冰盖留下了大量证据，证明它们以冰川地貌的形式表现出动态行为。在天气预报等领域，模型的改进是通过后向预测实现的--模拟过去的条件并将其与历史数据进行比较的行为。然而，对古冰盖的模拟很少充分利用冰川地貌记录。这个项目是一次刻意的尝试，目的是将冰盖建模和冰川地貌学这两个截然不同的学科结合在一起，以改进事后预测程序。在这个动机下，我的目标是开发和应用工具来验证冰盖模型与冰川地貌。在上一次冰川期间，格陵兰和南极的冰盖更大，北美、西北欧和巴塔哥尼亚的大部分地区都成了冰川。关于这些主要冰盖行为的信息目前以不同的方式存储。我将把覆盖这五个地区的冰盖的信息整理到一个标准化的数据库中。由此，我将对上一次冰川期间的冰盖范围和行为进行第一次全球范围的经验性重建。我还将使用最先进的冰盖模型来模拟这些冰盖的行为。冰盖将被模拟数百次，每次模拟都有一组单独的输入值，以捕捉所有可能的冰形态，考虑到我们目前关于冰盖如何流动和过去气候条件的知识有限。为了评估我的模型模拟，我将开发一套新的工具，它将量化模型对从冰川地貌数据集推断的古冰盖特性的预测能力。为了过滤掉极其不现实的模型模拟，我将首先将模型输出与我重建的冰盖范围进行比较。然后，根据冰川地貌记录的时间、流向、流动模式的变化、边缘位置、侵蚀能力和基本压力梯度的再现能力，计算模型性能的总体分数。然后，我将结合最符合地貌数据的模拟，对上一次冰川周期中存在的主要冰盖的行为进行高分辨率模拟。模拟的冰盖将作为现有冰盖的重要类似物，展示对地形、海平面和气候等冰盖行为的控制如何使这些冰盖退缩，以及它们不断变化的几何形状如何促成过去海平面的快速上升，例如融水脉冲和重大冰山排放事件。我的模型数据比较程序还将确定模型难以复制古冰盖行为的区域。这将突出模型中的不足之处，为模型开发提供目标。此外，未来南极和格陵兰冰盖的模型依赖于准确模拟它们过去的范围和行为。我进行的程序和模型实验将改进冰盖模型遵守这一历史的方式。因此，我的项目将带来冰盖模型测试和校准方式的阶段性变化，并为其他研究人员提供一个新的框架。

The ice sheets in Greenland and Antarctica hold enough water to raise global sea level by ~65 m, and are currently losing mass at an accelerating rate. Predictions of future mass loss are difficult to make, meaning that the contribution of ice sheets is a large source of uncertainty for estimates of future sea level rise. The numerical models which make these predictions need observations of ice sheet behaviour for validation and improvement, but the record of extant ice sheets is short. A longer record of ice sheet behaviour can be obtained by studying the regions occupied by ice sheets during the last glacial period (~120,000 to ~10,000 years ago). The ice sheets which existed during this time left behind abundant evidence for their dynamic behaviour in the form of glacial landforms. In fields such as weather forecasting, model improvement has been achieved by hindcasting - the act of simulating past conditions and comparing them to historical data. However, simulations of palaeo-ice sheets rarely utilise the glacial landform record to its full extent. This project is a deliberate attempt to bring together the rather disparate disciplines of ice sheet modelling and glacial geomorphology to improve hindcasting procedures. With this motivation, my aim is to develop and apply tools for validating ice sheet models with glacial geomorphology.During the last glacial, the Greenland and Antarctic ice sheets were larger and much of North America, Northwest Europe and Patagonia was glaciated. Information regarding the behaviour of these major ice sheets is currently disparately stored. I will collate information regarding the ice sheets that covered these five regions into a single standardised database. From this, I will produce the first global scale empirical reconstruction of ice sheet extent and behaviour during the last glacial. I will also use a state of the art ice sheet model to simulate the behaviour of these ice sheets. The ice sheets will be simulated hundreds of times, each simulation with a separate set of input values, to capture the full range of likely ice configurations given the current limitations of our knowledge regarding how ice sheets flow and past climate conditions.To evaluate my model simulations, I will develop a new set of tools which will quantify how well the model predicts the properties of palaeo-ice sheets inferred from the dataset of glacial geomorphology. To filter out extremely unrealistic model simulations, I will first compare model output to my reconstructed ice sheet extent. An overall score of model performance will then be calculated from the ability to reproduce the timing, flow direction, changes in flow pattern, margin position, erosive capabilities and basal pressure gradients that are recorded by glacial geomorphology.I will then combine the simulations which best conform to the geomorphological data to produce a high resolution simulation of the behaviour of the major ice sheets which existed during the last glacial cycle. The simulated ice sheets will serve as important analogues for existing ice sheets, demonstrating how controls on ice sheet behaviour such as topography, sea level and climate conditioned these ice sheets to retreat, and how their changing geometries contributed to past rapid sea level rise, such as meltwater pulses and major iceberg discharge events. My model-data comparison procedure will also identify regions where the model struggles to replicate the palaeo-ice sheet behaviour. This will highlight deficiencies in the model, providing targets for model development. Furthermore, models of the future Antarctic and Greenland ice sheets rely upon accurately simulating their past extent and behaviour. The procedures and model experiments I conduct will improve how ice sheet models adhere to this history. Therefore, my project will bring about a step-change in how ice sheet models are tested and calibrated and offer a new framework for other researchers to utilise.