Quantifying temporal and spatial causalities between climate change and slope failures

这个高风险研究项目的目标是开发数据驱动的方法工具，以确定影响或触发因素和指示信号，这些因素和指示信号表征了斜坡不稳定的早期阶段，这些不稳定会升级为山体滑坡、泥石流和崩塌。这些地质灾害通过破坏、破坏基础设施，甚至造成生命损失，对经济和公众造成负面影响。预期的项目成果是从特定的连续监测数据流中及时预测滑坡位移的可行性，为滑坡预警系统提供基础。传统的斜坡监测方法严重依赖地面观测(航空、无人机、卫星图像和全球定位系统数据)。一旦发生山体滑坡，就有大量的工作要做，早期还试图从实地调查的多尺度空间信息和区域尺度的航空/卫星数据中找出易发生山体滑坡的地点，即易感性评估。然而，由于地下记录的前兆信号尚未被完全理解或量化，因此及时预测斜坡尺度上迫在眉睫的滑坡仍然是一个具有挑战性的问题。从分离的软土体向山坡下移的地震信号的产生和记录已经被记录在案，但这种破坏的前兆信号的存在只有在实验室中才被证明。现场存在的前兆信号已被证明是岩石破坏的前兆信号，即裂缝的形成和扩展形状。据我们所知，目前还没有关于软土这类事件的公开目录/标签。我们假设软土崩塌确实会产生地震信号，这些地震信号可以被地震仪记录下来，并通过先进的信号处理来识别，但还没有找到完全支持这一说法的证据。为了确定在多大程度上可以及早发现和表征斜坡不稳定性，这个项目将调查滑坡的前兆和地下过程。我们将量化仪器/传感器模式、密度/粒度和山坡周围的地理区域，并结合先进的信号信息处理和机器学习(传统上将仪器和先进的分析分开处理)，以确定有效的实时预警系统的可行性。这种方法将从根本上改变我们对降水、温度和滑坡引发的地震活动之间的时间和空间因果关系的有限理解。目前的气候模型(例如，UKCIP)预测，由于气候变化，冬季更潮湿，降雨强度更大，英国气象局与英国气象局的合作表明，在暴雨期间，山体滑坡的数量显著增加。了解这些因果关系将有助于将新的研究领域发展为数据驱动的工程解决方案，以(I)从大片、有噪音和连续的记录中准确地提取地震预报信号，(Ii)将地面观测滑坡的仪器、测量地下地震活动的仪器与调查地下地震活动的地球物理方法联系起来，(Iii)扩大气候影响方案(例如，UKCIP)，以包括对滑坡的影响，(Iv)预测即将发生的滑坡及其规模。最终，这些将使我们能够减轻斜坡失稳对人类和经济的破坏性影响。

The goal of this high-risk research project is to develop data-driven methodological tools to identify influencing or triggering factors and indicator signals that characterise the early stages of slope instabilities, which escalate to landslide displacements, debris flow and rockfalls. These geological hazards negatively impact the economy and public, through disruption, damage of infrastructure and even loss of life. The intended project outcome is feasibility of timely landslide displacement prediction from particular continuous monitoring data streams, providing the basis for landslide early warning systems. Conventional approaches to slope monitoring rely heavily on surface observations (aerial, UAV and satellite image and GPS data). There is a large volume of work on detecting landslides once they have happened and there are early attempts at identifying locations prone to landslides, i.e., susceptibility assessments, from multi-scale spatial information from field surveys and aerial/satellite data at the catchment-to-regional-scale. However, timely prediction of imminent landslides at the slope-scale remains a challenging problem because precursory signals from subsurface recordings are as yet not fully understood or quantified.The generation and recording of seismic signals from a detached soft soil mass that is moving downwards a mountain slope has been documented, but the presence of precursory signals for such failures has only been shown in the lab. The presence of precursory signals in the field has been documented for rock failure, i.e., in the shape of the formation and propagation of cracks. We know of no publicly available catalogues/labels of such events for soft soils. We hypothesise that soft soil failure does generate seismic signals that can be recorded by seismometers and identified through advanced signal processing but the evidence to fully support this statement is yet to be found.To determine to what extent early detection and characterisation of slope instabilities is possible, this project will investigate the precursors to a landslide and the underlying subsurface processes. We will quantify the instrumentation/sensor modalities, density/granularity and geographic area around a hill slope, in conjunction with advanced signal information processing and machine learning (instrumentation and advanced analysis are traditionally treated in isolation), to determine the feasibility of an effective real-time warning system. This approach will radically transform our very limited understanding of temporal and spatial causalities between precipitation, temperature, and landslide induced seismicity. Current climate modelling (e.g., UKCIP) is predicting wetter winters and higher intensity of rainfall due to climate change, and the Met Office with BGS have demonstrated a marked increase in the number of landslides at times of heavy rainfall. Understanding these causalities will enable the development of new fields of research into data-driven engineering solutions to (i) accurately extract seismic predictor signals from large, noisy and continuous recordings, (ii) make linkages between instrumentation that make surface observations of landslides, measures seismicity at subsurface and geophysical approaches that interrogate the subsurface, (iii) augment climate impact programme (e.g., UKCIP) to include effect on landslides, (iv) predict an impending landslide and its scale. Ultimately, these will enable us mitigate the devastating effect of slope instabilities on humans and the economy.