GLOBE: The Greenland Subglacial Lake Observatory

理论预测表明，格陵兰冰盖下隐藏着1600个冰下湖泊。然而，到目前为止，只有不到5%的人被发现，研究发现只有7人正在积极填充或排出。追踪格陵兰岛下的活跃湖泊是有问题的，因为它们比南极洲下面的湖泊小100倍，因此大多数标准的卫星方法都无法检测到它们。在这里，我将从根本上提高探测、监测、预测和了解规模上活跃的格陵兰冰下湖泊动态的能力。Global将把这一领域从目前几乎对其分布、动态、驱动因素和影响一无所知的状态转变为一个对所有湖泊进行实时监测、解决它们的历史、预测它们未来排水的位置。Global将通过三个关键领域的创新实现这一目标，1)大规模利用新的超高分辨率卫星数据，2)开发利用非常规数据的新方法，3)开发新的深度学习和统计预测方法。具体地说，我将1)使用高容量的数字高程模型和机器学习来首次绘制整个大陆以及过去25年的湖泊活动图，2)训练深度学习模型以在历史卫星图像中搜索活跃的湖泊，3)使用视频监控方法自主识别发生的新的湖泊活动，以及4)使用统计技术对未来湖泊演变作出初步预测。最终，GLOBAL将推动一种全新的大规模湖泊监测方法，创建一个自主跟踪、预测和传播湖泊活动的格陵兰冰下湖泊观测站，从而改变对难以捉摸的湖泊排水事件的驱动因素、动态和影响的理解。

Theoretical predictions suggest that 1600 subglacial lakes lie hidden beneath the Greenland Ice Sheet. Yet less than 5% have been discovered to date, and studies have found only 7 that are actively filling or draining. Tracking active lakes beneath Greenland is problematic because they are 100 times smaller than those beneath Antarctica, rendering them undetectable by most standard satellite methods. Here, I will radically advance the ability to detect, monitor, forecast and understand the dynamics of active Greenland subglacial lakes at scale. GLOBE will transform the field from its current state, where almost nothing is known about their distribution, dynamics, drivers and impacts; to a position where all lakes are monitored in real time, their histories are resolved, and probabilistic forecasts predict their future drainage. GLOBE will achieve this through innovation in three key areas, 1) exploiting new, super-high resolution satellite data at scale, 2) developing new methods that exploit unconventional data, and 3) developing novel deep learning and statistical forecasting methodologies. Specifically, I will 1) use high-volume digital elevation models and machine learning to map lake activity, for the first time, over the entire continent and the past 25 years, 2) train deep learning models to search for active lakes in historical satellite imagery, 3) use video surveillance methods to autonomously identify new lake activity as it happens, and 4) use statistical techniques to make the first forecasts of future lake evolution. Ultimately, GLOBE will catalyse an entirely new approach to large-scale lake monitoring, creating a Greenland Subglacial Lake Observatory that autonomously tracks, forecasts and disseminates lake activity, thus transforming understanding of the drivers, dynamics and impacts of elusive lake drainage events.