Interpretable Machine Learning Modelling of Future Extreme Floods under Climate Change

机器学习(ML)是一种强大的技术，它迅速地在许多学科中找到了应用。估计气候变化对未来极端事件，如超级洪水和热浪的影响，传统上是使用基于物理和基于过程的水文气候模型进行的；然而，基于ML的建模已经成为一个新的重点领域，因为它能够在这种背景下考虑高度复杂的、非线性的物理和社会-物理相互作用。ML对未来极端洪水的建模仍然存在一个巨大的挑战，那就是它的黑箱性质，各个组件之间的相互作用几乎无法解释，这阻碍了它在支持关键决策过程以缓解气候变化方面的有效应用。本博士项目旨在通过开发一套具有更好的可解释性和透明度的ML建模工具包来应对这些挑战，以预测气候变化影响下的未来极端洪水的风险。该项目的预期成果将为气候变化下针对极端洪水的缓解措施的关键决策提供负责任和可靠的模型支持。参与的博士生将得到一个经验丰富的监督团队的支持，该团队由学者、行业合作伙伴以及具有基于HPC的水文气候建模、人工智能和机器学习专业知识的一线从业人员组成。这名博士生还将受益于东道主大学通过齐恩凯维奇研究所提供的建模、人工智能和职业发展方面的系统培训，主题是气候行动中的建模、数据和人工智能。

Machine learning (ML) is a powerful technique that has rapidly found many applications across numerous disciplines. Estimating climate change impact on future extreme events such as super floods and heatwaves has been traditionally carried out using physically and process-based hydro-climatic models; however, ML-based modelling has become a new focal area thanks to its ability to account for the highly complex, nonlinear physical and social-physical interactions in this context. A huge challenge that still remains in ML modelling for future extreme floods, is its 'black-box' nature where the interactions among various components are hardly explainable, which hinders its effective application in supporting critical decision-making processes to mitigating climate change.This PhD project aims to address these challenges by developing a set of ML modelling tool kits with improved interpretability and transparency for predicting future risks of extreme floods under the impact of climate change. The expected outcome of the project will offer an accountable and reliable modelling support for critical decision making on mitigation measures against extreme floods under the changing climate. The PhD student involved will be supported by an experienced supervision team consisting of academics, industrial partners as well as frontline practitioners with expertise in HPC-based hydro-climatic modelling, AI and machine learning. The PhD student will also be benefitted from the systematic training in modelling and AI, career development offered by the host university via it's the Zienkiewicz Research Institute under the theme of Modelling, Data and AI in Climate Action.