Towards reliable assessment of pyroclastic density current hazards

在一个人口不断增长并被迫生活在危险的火山地区的世界里，火山正日益成为对生命的更大威胁。火山爆发对社会产生了毁灭性的影响，包括将国家覆盖在灰烬中，改变气候，以及大量人类生命的损失。最严重的火山灾害是由火山流引起的，其中包括滑坡、泥石流和最危险的火山碎屑流。由灼热的火山灰和气体组成的火山碎屑流可以在几分钟内烧毁并掩埋整个城市。这些热流通常由基底密集的雪崩和上覆的稀释湍流火山灰云浪涌组成。火山碎屑流不仅会影响地面，还会形成大量细小的火山灰颗粒，上升到巡航航班的高度，扰乱航空路线。预测这些流体的传播一直是地球科学中最大的挑战之一，因为我们对复杂颗粒介质的流动缺乏基本的了解，也就是说，我们对它们的流变性的了解非常不完整。我们知识上的这一差距使得火山碎屑流的影响很难预测。预测未来火山碎屑流速度和淹没区域的能力将有助于限制人员生命损失，并通过通知疏散等减灾战略减少经济影响。不幸的是，在我们捕捉到这些电流的物理特性并在数值模型中实现之前，这个目标是无法实现的。密集的雪崩层是一种高度复杂的颗粒流，由大小不等（从微米到米）的颗粒组成。气-颗粒耦合导致气体压力升高，并使高摩擦颗粒雪崩转变为类似液体的混合物。虽然我们对颗粒流动的理解在过去十年中有了显著的增长，但以前的研究主要集中在稳定的结构和简化的颗粒混合物上。在自然界中，火山碎屑流随着时间的推移而演变，因为颗粒相互碰撞而破碎和磨损，并且流动穿过各种地形障碍，如控制其行为的山谷，使其行为短暂。如果没有对天然火山混合物的不稳定流变学的物理描述，我们可能永远无法准确地捕捉它们的行为。我们面临的另一个主要挑战是，目前的模型需要在高分辨率的数字高程模型上运行火山碎屑流模拟。目前，所有的模型都使用中央处理器（CPU）来模拟火山流，需要超级计算机来解决数百个场景，需要几天到几周的时间才能完成。该项目将利用物理学和工程学中可用的计算能力和分析技术的最新进展，并将其应用于地球科学。这些技术将用于研究非定常火山碎屑混合物的耗散能，使物理描述过程能够在基于图形卡的新一代火山流模型中实现。这款新机型将使用图形处理单元（GPU）计算，可以在任何笔记本电脑上进行。这个新模型将允许高分辨率的计算，并将从根本上改变我们预测火山碎屑流危害及其与地形的相互作用的能力，并使火山学家能够在最需要的时候进行快速的危害评估：在火山动荡期间。将这项研究的发现和发展与地球科学的其他领域相结合，将在如何进行火山危害评估方面取得重要进展，并有助于限制生命损失。

In a world where the human population keeps growing and is pushed to living in hazardous volcanic areas, volcanoes are increasingly becoming a larger threat to life. Volcanoes that erupt explosively have had devastating societal impacts, including covering countries in ash, changing the climate, and the extensive loss of human life. The most serious class of volcanic hazards is caused by volcanic flows, which include landslides, debris flows and the most dangerous of all, pyroclastic flows. Pyroclastic flows, made of scorching ash and gas, can burn and bury entire cities within minutes. These hot currents are typically composed of a basal dense avalanche and overriding dilute turbulent ash-cloud surge. Pyroclastic flows do not solely affect the ground, as they can also form large plumes of fine ash particles that rise to the altitude of cruising flights and can disrupt aviation paths. Predicting the propagation of these flows has been one of the largest challenges in geosciences because we lack a fundamental understanding of how complex granular media flow, i.e. our understanding of their rheology is very incomplete. This gap in our knowledge makes the impacts from pyroclastic flows very difficult to predict. The ability to forecast future pyroclastic flow velocity and inundation areas would help to limit the loss of human life and reduce economic impacts by informing mitigation strategies such as evacuations. Unfortunately, this goal cannot be achieved until we capture the physics of these currents and implement it in numerical models. The dense avalanche layer is a highly complex granular flow made of particles spanning a wide range of sizes (from microns to meters). The gas-particle coupling leads to elevated gas pressure and enables the transformation of the highly frictional granular avalanche into a mixture analogous to a liquid. While our understanding of granular flows has grown significantly in the past decade, previous studies have focused on steady configurations and simplified mixtures of grains. In nature, pyroclastic flows evolve over time as particles fragment and abrade by colliding with each other, and flows propagate across a variety of topographic obstacles such as valleys that control their behaviour, making their behaviour transient. Without a physical description of unsteady rheology of natural volcanic mixtures, we may never capture their behaviour accurately. Another major challenge we face is the time that current models require to run simulations of pyroclastic flows on highly resolved digital-elevation models. At the moment, all models use Central Processing Unit (CPU) computing to simulate volcanic flows, and require supercomputers to solve hundreds of scenarios taking days to weeks to complete. This project will take advantage of recent advances in computing abilities and analytical techniques available in physics and engineering and apply these to geosciences. These techniques will be used to study the dissipation energy from unsteady pyroclastic mixtures, enabling physical descriptions of the processes to be implemented in a new generation of volcanic flow model based on graphic cards. This new model will use Graphic Processing Unit (GPU) computing that can be undertaken on any laptop. This new model will allow highly resolved calculations and will radically transform our ability to forecast pyroclastic flow hazards and their interaction with topography, and enable volcanologists to undertake rapid hazard assessment when most needed: during volcanic unrest. Combining the findings and development from this study with other fields in geosciences will lead to important advances in how volcanic hazard assessment is undertaken and help limit loss of life.