EXCESS: The role of excess topography and peak ground acceleration on earthquake-preconditioning of landslides

滑坡是导致岩石和土壤破坏并沿斜坡向下移动的物理过程的总称。当陡坡因暴雨、地震、自然过程（例如河流）冲刷坡底或人为活动导致山坡上的物质崩塌等因素而失去稳定时，就会发生滑坡。全球每年发生数千次山体滑坡，造成数千人死亡（例如，从2004年到2016年，有55997人死于4862次山体滑坡事件），并严重破坏基础设施，扰乱经济，阻碍国际发展。尽管进行了广泛的研究，但预测滑坡何时何地发生的能力仍然是一项基本的科学挑战。这在一定程度上是因为科学家们认为某一地区的滑坡率每年都是恒定的，而且滑坡会发生在这些景观的类似地方。如果是这样的话，那么就可以很容易地了解最可能发生山体滑坡的地点和时间，也就是说，它们是“可预测的”。不幸的是，最近的研究表明，这种假设是不正确的，事实上，风暴和地震等突发的极端事件会改变山体滑坡的速度和模式。因此，能够预测滑坡风险高的地区仍然是灾害管理的重要前沿。地震不仅会引起山体滑坡，因为地震期间地面变形和震动，而且在地震发生后，在接下来的1-10年里，山体滑坡的数量会增加——这个过程被称为“地震预处理”。这种现象造成了一种额外的危害和风险，而这种危害和风险在很大程度上没有得到认识和量化。我们最近在尼泊尔进行的突破性研究表明，地震强度与过量地形（地形中高于稳定阈值斜坡的区域）和随后的山体滑坡之间存在联系。如果这种关系在世界其他地区也是如此，我们将有一种高度创新的方法来定位风险较高的地区。该项目将通过对近期事件和计算机建模的研究来解决这一关键的研究前沿。首先，我们将使用高分辨率（<5米）卫星图像，在六个不同地区的最近大地震之前、期间和之后创建新的滑坡目录。这些高分辨率的数据使我们能够准确地确定每个地区滑坡发生的长期平均率，并自信地确定地震后滑坡增加的规模和持续时间。所选择的地区和地震跨越一系列气候，构造环境和地震规模，使我们能够调查影响，并确定不同控制因素（例如，降雨量，坡度，地形，地震规模）在全球范围内的相对重要性，确保研究成果具有广泛的适用性。然后，这些数据集将用于区域一级的滑坡易感性模型，形成可供国家/区域政府和机构用于减轻灾害和风险的产出。其次，我们将开发一个新的基于过程的计算机模型来研究地震景观破坏的机制，以及这种破坏如何随着时间的推移而变化，从而导致观测到的滑坡模式。与经验统计模型不同，基于过程的模型明确地模拟了滑坡发生的驱动因素，并可以考虑突然和快速的环境变化的影响。该模型的结果将通过敏感性图进行验证，并且能够模拟10年至1000年的多次地震，这将使人们对地震诱发和地震预处理的滑坡在长期景观演变中的作用有新的认识，最终提高准确预测地震周期中滑坡位置的能力。

Landsliding is a collective term for physical processes that cause rock and soil to fail and move down slope. Landslides occur when steep slopes are destabilised by factors such as heavy rainfall, earthquakes, the removal of the base of the slope by natural processes (e.g., by rivers) or by the action of people causing material on the hillside to collapse. Many thousands of landslides occur globally each year, killing thousands of people (e.g., from 2004 and 2016; 55,997 people died in 4,862 separate landslide events) and significantly damaging infrastructure, disrupting economies and hindering international development. Despite extensive research, the ability to forecast when and where a landslide will occur remains a fundamental scientific challenge. This is partly because scientists had thought that the rate of landsliding in a certain area is constant from year to year, and that landslides would occur in similar places in those landscapes. If this were the case, then it would be straightforward to understand where and when landslides would most likely occur, i.e., they would be 'predictable'. Unfortunately, recent research shows that such assumptions are incorrect and in fact sudden extreme events such as storms and earthquakes will change the rates and patterns of landsliding. Being able to predict areas of elevated landslide risk thus remains an imperative frontier in hazard management. Earthquakes not only induce landslides because of ground deformation and shaking during the event, but also after an earthquake there are increased numbers of subsequent landslides over the next 1-10 years - this process has been termed "earthquake-preconditioning". This phenomenon poses an additional hazard and risk that is largely unrecognised and unquantified. Our recent ground-breaking research in Nepal suggests that there is a link between the strength of an earthquake and excess topography (areas in the landscape that are above a stable threshold slope) and subsequent landsliding. If this relationship is true in other parts of the world, we will have a highly innovative way of locating areas at higher risk.This project will address this critical research frontier through the study of recent events and computer modelling. Firstly, we will create new landslide catalogues before, during and after recent large earthquakes for six different regions, using high-resolution (<5m) satellite imagery. These high-resolution data allow us to accurately determine the long-term average rate of landslide occurrence in each region and confidently identify the size and duration of periods of increased landsliding following an earthquake. The regions and earthquakes selected span a range of climates, tectonic settings, and earthquake sizes to enable us to investigate the influence, and determine the relative importance that different control factors (e.g., rainfall, slope, topography, earthquake size) have at a global level, ensuring that the research outputs have wide applicability. These datasets will then be used in landslide susceptibility models at regional level to form outputs that can be used in hazard and risk mitigation by national/regional governments and agencies. Secondly, we will develop a new process-based computer model to investigate the mechanism of earthquake landscape damage and how this changes through time to cause observed patterns of landslides. Unlike empirical statistical models, process-based models explicitly simulate the drivers of landslide occurrence and can consider the impact of sudden and rapid environmental changes. The results of the model will be validated by the susceptibility maps, and the ability to model multiple earthquakes over 10s to 1000s of years will lead to new insights into the role of earthquake-induced and earthquake-preconditioned landslides in long-term landscape evolution, ultimately increasing the ability to accurately forecast the location of landslides across earthquake cycles.