Improving the Predictivity of Simulating Natural Hazards due to Mass Movements – Optimal Design and Model Selection –

提高模拟大规模运动造成的自然灾害的预测能力 â 优化设计和模型选择 â

• 批准号: 441527981
• 负责人: Professorin Dr. Julia Kowalski
• 金额: --
• 依托单位: Lehrstuhl für Methoden der Modellbasierten Entwicklung in den Computergestützten Ingenieurwissenschaften
• 依托单位国家: 德国
• 项目类别: Research Grants
• 资助国家: 德国
• 项目状态: 未结题
• 来源: https://gepris.dfg.de/gepris/projekt/441527981?language=en
• 关键词: Improving; Predictivity; Simulating; Natural; Hazards

The goal of this project is to improve the predictive power of computational process models for rapid gravity-driven mass movements such as debris flows or other forms of complex landslides as well as rock or snow avalanches. These computational models are used to compute run-out distances, deposition areas and impact pressures, all of which are important when wanting to design hazard mitigation measures. Mass movements are defined as the movement of mobilized surface material caused by gravity. Their dynamic behaviour is highly influenced by their flowing material. Its composition varies depending on its exact type (landslide versus debris flow versus snow avalanche). It is often highly heterogeneous and, due to entrainment and deposition along the way, might even vary with time. Especially the amount of water in the flowing body can have a dramatic impact on the mass movement’s dynamic behaviour. A major challenge when developing computational models for gravity-driven mass movements is hence to adequately reflect material and process complexity in the underlying mathematical process model. To date, research on improving the computational model's predictive power mainly focuses on developing process models of higher complexity levels. Instead of treating the mobilized mass as one bulk mixture, for example, water and solids of different sizes are modeled separately, or the dimensionality of the modeling approach is increased. While such strategies can offer new insights into the process itself, they typically come at the price of an increasing number of model parameters, which are hard to calibrate.Another possible strategy to improve the predictive power is often overlooked, namely to systemically optimize the information content of the data sets used for calibrating model parameters, as well as to exploit the available data sets in a better way. We will address this research gap in the proposed research project by developing optimal design concepts that will result in novel, optimized field scale monitoring set-ups and protocols as well as data sets of higher information content. In a second phase of the project, we will develop model selection routines that allow to utilize available data sets for gravity-driven mass movements to find the most plausible process model out of a collection of candidate models. These two aspects are novel to the field of natural hazards research. They will be developed and validated by a doctoral student, who will be supervised by the applicant in her recently granted Heisenberg group. The value-add of the developed methodologies for the natural hazards engineering community will be demonstrated during a number of national and international collaborations.

该项目的目标是提高计算过程模型对重力驱动的快速质量运动（如泥石流或其他形式的复杂滑坡以及岩石或雪崩）的预测能力。这些计算模型用于计算跳动距离、沉积面积和冲击压力，所有这些在设计减灾措施时都很重要。质量运动被定义为由重力引起的移动的表面材料的运动。它们的动态行为受到其流动材料的高度影响。它的组成取决于它的确切类型（滑坡、泥石流和雪崩）。它往往是高度不均匀的，由于夹带和沉积沿着，甚至可能随时间而变化。特别是流动体中的水量会对质量运动的动态行为产生巨大影响。因此，在开发重力驱动的质量运动的计算模型时，一个主要的挑战是在底层的数学过程模型中充分反映材料和过程的复杂性。到目前为止，提高计算模型的预测能力的研究主要集中在开发更高的复杂程度的过程模型。例如，代替将活动化的物质作为一个散装混合物处理，不同尺寸的水和固体被单独建模，或者建模方法的维度被增加。虽然这些策略可以为过程本身提供新的见解，但它们通常以越来越多的模型参数为代价，这些参数很难校准。另一种可能的提高预测能力的策略经常被忽视，即系统地优化用于校准模型参数的数据集的信息内容，以及以更好的方式利用可用的数据集。我们将通过开发最佳设计概念来解决拟议研究项目中的这一研究空白，这些概念将导致新的，优化的现场规模监测设置和协议以及更高信息含量的数据集。在该项目的第二阶段，我们将开发模型选择例程，允许利用重力驱动的质量运动的可用数据集，以找到最合理的过程模型的候选模型的集合。这两个方面是新的自然灾害研究领域。它们将由一名博士生开发和验证，该博士生将由申请人在她最近获得的海森堡群中监督。将在一些国家和国际合作中展示所制定的方法对自然灾害工程界的增值作用。